JP5701010B2 - Image processing apparatus and control method thereof - Google Patents

Description

  The present invention relates to an image processing apparatus that performs processing by dividing an image into a plurality of regions.

  As the resolution of an image or a display device increases, there is a technique in which an image is divided into a plurality of regions and image correction is performed using a plurality of image processing units. Patent Document 1 below discloses a technique for performing image processing using total data such as a histogram detected by each image processing unit from each image divided into a plurality of regions. The statistical value calculation unit acquires a plurality of total data detected by the plurality of image processing units, and each image processing unit performs image processing based on the statistical values created from the acquired total data.

JP 2006-71938 A

  However, in Patent Document 1, the data transmitted from each image processing unit to the statistical value calculation unit is total data such as a histogram that is a value itself detected from each image, and therefore the amount of data to be transmitted is large. It takes time. When data transmission takes time, there is a problem that time for image processing is reduced, or image processing in real time becomes difficult.

  Therefore, in the present invention, the data transmission time is reduced by reducing the amount of data transmitted from each image processing unit that performs image processing of each image divided into a plurality of regions, and the time for performing image processing is secured. It aims to make it possible to do.

In order to achieve the above object, the image processing apparatus of the present invention includes a plurality of image processing unit for performing image processing for a plurality of divided images obtained by dividing an input image into a plurality of regions, each divided image Correction data creating means for creating correction data for each divided image for correcting the gradation of each of the plurality of divided images , and each image processing unit comprises a histogram creating means for creating a histogram in each of the plurality of divided images; , from the histogram, and the intermediate data generating means for generating intermediate data including the representative value of the gradation in each divided image, a boundary value of the gradation of a region which is in contact with the other divided image in each divided image, before Symbol have a, a correction processing means for performing processing for correcting the gradation of each of the divided image based on the correction data generated by the correction data forming means, the correction data generation means, said in It created by the data creation means, from the intermediate data in each of the plurality of divided images, characterized by creating correction data for each divided image in order to correct the gradation of each divided image.

  According to the present invention, the data transmission time is reduced by reducing the amount of data transmitted from each image processing unit that performs image processing of each image divided into a plurality of regions, and the time for performing image processing is ensured. can do.

1 is a block diagram of an image processing apparatus according to Embodiment 1. FIG. It is a figure which shows the relationship between an input image and a divided image. 1 is a block diagram of an image processing unit according to Embodiment 1. FIG. It is a figure which shows the relationship between a divided image and a unit area. It is a figure explaining the detection method of an attention gradation. It is a figure which shows the representative value of divided image Sa, and the boundary value of divided image Sa. 3 is a block diagram of a correction data creation unit according to Embodiment 1. FIG. It is a figure explaining the unit area | region of the boundary value for correction | amendment contained in the data for correction | amendment. It is a figure which shows the conversion characteristic of a full screen gamma curve. It is a figure explaining the process which calculates | requires the local correction value in the pixel P by bilinear interpolation. It is a figure which shows the conversion characteristic of a local (gamma) curve. FIG. 3 is a block diagram of an image processing apparatus according to a second embodiment. FIG. 6 is a block diagram of an image processing unit according to a second embodiment.

  Embodiments of the present invention will be described below with reference to the accompanying drawings.

Example 1
FIG. 1 is a block diagram of this embodiment. An image processing apparatus 100 illustrated in FIG. 1 includes an image dividing unit 102, image processing units 103a, 103b, 103c, and 103d, a correction data creating unit 104, and an image combining unit 105. In this embodiment, it is assumed that one image processing unit is mounted on one IC (Integrated Circuit). The operation of the image processing apparatus 100 configured as described above will be described below.

  The image dividing unit 102 divides the input image into a plurality of regions and creates a plurality of divided images. FIG. 2 is a diagram for explaining a method of dividing an input image by the image dividing unit 102. FIG. 2A shows an input image S input to the image dividing unit 102. In this embodiment, the resolution of the input image S is 3840 × 2160 pixels. FIG. 2B shows a divided image created by the image dividing unit 102. In this embodiment, the input image S is divided into four divided images Sa, Sb, Sc, and Sd. The divided images Sa, Sb, Sc, and Sd are output to the image processing units 103a, 103b, 103c, and 103d, respectively, and are subjected to image processing by each image processing unit. The resolution of the divided images Sa, Sb, Sc, and Sd is 1920 × 1080 pixels.

  Next, the image processing unit 103a will be described. Since the image processing units 103b, 103c, and 103d have the same configuration as the image processing unit 103a and perform the same processing, only the image processing unit 103a will be described in the present embodiment.

  FIG. 3 is a block diagram showing an internal configuration of the image processing unit 103a. The image processing unit 103a includes a histogram creation unit 10a, a target gradation detection unit 11a, an intermediate data creation unit 12a, a difference value calculation unit 13a, a full screen γ generation unit 14a, a local γ generation unit 15a, a γ correction processing unit 16a, an IF (Interface) part 17a is comprised. In addition, it is assumed that the target gradation detection unit 11a, the intermediate data creation unit 12a, the difference value calculation unit 13a, and the full screen γ generation unit 14a are inside the microcomputer 18a. The full screen γ generation unit 14a may be outside the microcomputer 18a.

  The histogram creation unit 10a detects a luminance histogram Ha (i) in each unit region obtained by dividing the divided image Sa. In the present embodiment, for example, as shown in FIG. 4, the divided image Sa is divided into 5 × 3 × 15 unit areas a0 to a14. Accordingly, fifteen luminance histograms Ha (0) to Ha (14) are detected from the histogram creation unit 10a. In the luminance histogram Ha (i), for example, the number of pixels of each gradation is counted with 256 gradations of 0 to 255. The luminance histograms Ha (0) to Ha (14) detected by the histogram creation unit 10a are output to the target gradation detection unit 11a.

  The attention gradation detecting unit 11a detects the dark part attention gradation Nba (i) and the bright part attention gradation Nwa (i) from each luminance histogram Ha (i). FIG. 5 is a diagram for explaining a method of detecting the dark part attention gradation Nba (i) and the bright part attention gradation Nwa (i) from the luminance histogram Ha (i).

  The target gradation in the luminance histogram Ha (i) is a gradation value for which it is desirable to improve gradation. The target gradation is (1) a gradation value in which the frequency is equal to or greater than a predetermined threshold and (2) a maximum value in the luminance histogram Ha (i), and (3) a frequency within a predetermined range including the gradation value. The variation value is a gradation value smaller than a predetermined reference, and is a gradation value satisfying the above three conditions (1) to (3). Here, whether or not the frequency under the condition (2) is a maximum value is, for example, the frequency of the target gradation value whether or not it is the target gradation and m gradation values before and after that (m is 1). What is necessary is just to judge by comparing with the frequency of the above integer). The maximum value among the frequencies of the gradation values to be compared is set as the maximum value. Note that m tone values before and after the target tone value are m tone values smaller than the target tone value and m levels larger than the target tone value. Means the key value. The condition (3) is that when the image includes an area of the same color (same gradation value) (for example, a letterbox or a graphic or character such as data broadcasting), the frequency variation amount is locally increased. This is a condition for excluding key values. Whether or not the condition of (3) is satisfied is determined by the frequency of the target gradation value as to whether or not it is the target gradation, and the gradation value of ± n excluding the gradation value (n is an integer of 1 or more) The determination may be made by comparing with a value obtained by multiplying the sum of frequencies by a predetermined value.

  The dark part attention gradation in the luminance histogram Ha (i) is an attention gradation whose gradation value is on the low gradation side. In this embodiment, when the luminance histogram is 256 gradations, the minimum attention gradation among the attention gradations satisfying the above condition in the range of gradation values from 2 to 126 is set as the dark part attention gradation. When there is no attention gradation in the range of gradation values 2 to 126, the dark part attention gradation is set to 128. The bright part attention gradations in the luminance histogram Ha (i) are attention gradations whose gradation values are on the higher gradation side. In this embodiment, when the luminance histogram is 256 gradations, the maximum attention gradation among the attention gradations satisfying the above condition in the gradation value range of 130 to 253 is set as the bright part attention gradation. When there is no target gradation in the range of gradation values 130 to 253, the bright part attention gradation is set to 128. In the gradation values 0, 1, 127, 128, 129, 254, and 255, the frequencies of the two upper and lower gradation values are excluded from the attention gradation range in order to use the attention gradation detection.

  When the divided image Sa is divided into unit areas as shown in FIG. 4, the attention gradation detecting unit 11a determines the dark part attention gradations Nba (0) to Nba (14) and the luminance histograms Ha (0) to Ha (14) and Bright part attention gradations Nwa (0) to Nwa (14) are detected. The detected dark part attention gradations Nba (0) to Nba (14) and bright part attention gradations Nwa (0) to Nwa (14) are output to the intermediate data creation unit 12a and the difference value calculation unit 13a.

  The intermediate data creation unit 12a creates the intermediate data Ma of the divided image Sa from the dark part attention gradations Nba (0) to Nba (14) and the bright part attention gradations Nwa (0) to Nwa (14). The intermediate data Ma of the divided image Sa is data including a representative value and a boundary value of the divided image Sa described below.

The representative values are the minimum value mina of the plurality of dark part attention gradations Nba (0) to Nba (14) and the maximum value of the plurality of bright part attention gradations Nwa (0) to Nwa (14). It consists of maxa. When the divided image Sa is divided into unit areas as shown in FIG. 4, the minimum value mina and the maximum value maxa are obtained by the following expressions 1 and 2.
mina = min [Nba (0) to Hba (14)] (Formula 1)
min []: Minimum value in [] maxa = max [Nwa (0) to Nwa (14)] (Formula 2)
max []: Maximum value in [] On the other hand, the boundary value is the dark part attention gradation Nba (i) and the bright part attention of the unit area in contact with the other divided images Sb, Sc, Sd in the divided image Sa. This is the gradation Nwa (i).

  The representative value and boundary value of the divided image Sa will be described with reference to FIG. FIG. 6 is an example showing the dark part attention gradation Nba (i) and the bright part attention gradation Nwa (i) of each unit area of the divided image Sa divided into the unit areas a0 to a14 as shown in FIG. . In the example of FIG. 6A, the minimum value mina of the 15 dark part attention gradations Nba (0) to Nba (14) is 36. In FIG. 6B, the maximum value maxa of the 15 bright part attention gradations Nwa (0) to Nwa (14) is 225. From the above, the minimum value mina = 36 and the maximum value maxa = 225 are the representative values of the divided image Sa.

  As shown in FIG. 2B, the divided image Sa has a divided image Sb on the right, a divided image Sd on the lower right, and a divided image Sc on the lower side. The unit areas a4, a9 to. a14 is in contact with the other divided images Sb, Sc, and Sd. Therefore, the boundary value of the divided image Sa is composed of the dark part attention gradation Nba (i) and the bright part attention gradation Nwa (i) of each of the unit regions a4 and a9 to a14. When the input image is composed of the divided image Sa and the divided image Sb (when the input image is divided into two and the divided image Sb is on the right of the divided image Sa), the boundary values of the divided image Sa are unit regions a4, a9, Each of a14 includes a dark part attention gradation Nba (i) and a bright part attention gradation Nwa (i).

  The intermediate data Ma of the divided image Sa made up of the representative value of the divided image Sa and the boundary value of the divided image Sa created by the intermediate data creating unit 12a is output to the correction data creating unit 104 through the IF unit 17a.

  The correction data creation unit 104 obtains correction data Ua, Ub, Uc, Ud from the intermediate data Ma, Mb, Mc, Md created by the intermediate data creation unit of each of the image processing units 103a, 103b, 103c, 103d. create. The intermediate data Mb is composed of the representative values (minb, maxb) and boundary values of the divided image Sb, the intermediate data Mc is composed of the representative values (minc, maxc) and boundary values of the divided image Sc, and the intermediate data Md is the divided image Sd. It consists of representative values (mind, maxd) and boundary values. The operation in the correction data creation unit 104 will be described with reference to FIG.

  In this embodiment, the correction data Ua, Ub, Uc, and Ud each include a full-screen correction value and a correction boundary value described below. The full screen correction value includes a full screen dark portion value B and a full screen bright portion value W, and the full screen dark portion value B and the full screen bright portion of the full screen correction values included in the correction data Ua, Ub, Uc, Ud, respectively. The value W is the same value and a common value. The boundary value for correction is a dark part attention gradation and a bright part attention gradation in a unit region in another divided image in contact with an arbitrary divided image.

  In the correction data creation unit 104, the intermediate data receiving unit 71 receives the intermediate data Ma, Mb, Mc, and Md transmitted from the image processing units 103a, 103b, 103c, and 103d. The separation unit 72 separates the representative value and the boundary value from each intermediate data, transmits the representative value of each intermediate data to the full-screen correction value calculation unit 73, and the boundary value of each intermediate data is determined as the correction boundary value determination unit 74. Send to.

The full screen correction value calculation unit 73 calculates a full screen correction value including the full screen dark part value B and the full screen bright part value W. The full screen dark area value B is the minimum value among the minimum values mina, minb, minc, and mind among the representative values of the divided images included in the intermediate data Ma, Mb, Mc, and Md. The full screen bright part value W is the maximum value among the maximum values maxa, maxb, maxc, and maxd among the representative values of each divided image. That is, the full screen dark part value B and the full screen bright part value W are obtained by the following equations 3 and 4.
B = min [mina, minb, minc, mind] (Formula 3)
W = max [maxa, maxb, maxc, maxd] (Formula 4)

  The correction boundary value included in the correction data will be described using the correction data Ua as an example. The correction boundary values included in the correction data Ua are the dark part attention gradation and the bright part attention gradation in the unit area of another divided image in contact with the divided image Sa. FIG. 8 is a diagram illustrating the correction boundary value included in the correction data Ua. The unit areas of the other divided images in contact with the divided image Sa are b0, b5, and b10 of the divided image Sb shown in gray in FIG. 8, c0 to c4 of the divided image Sc, and d0 of the divided image Sd. Therefore, the boundary values for correction of the divided image Sa included in the correction data Ua are the dark part attention gradation and the bright part attention gradation in each of the unit areas b0, b5, b10, c0 to c4, and d0. When the input image is composed of the divided image Sa and the divided image Sb, the correction boundary values of the divided image Sa included in the correction data Ua are the dark part attention gradation and the bright part attention gradation in each of the unit regions b0, b5, and b10. It is.

  As described above, the correction boundary value of the divided image Sa included in the correction data Ua is determined from the boundary values transmitted from the separation unit 72 by the correction boundary value determination unit 74. In the same process, the correction boundary value of the divided image Sb included in the correction data Ub, the correction boundary value of the divided image Sc included in the correction data Uc, and the correction of the divided image Sd included in the correction data Ud. Each boundary value is determined.

  The correction data output unit corrects each divided image by using the full screen correction value calculated by the full screen correction value calculation unit 73 and the correction boundary value of each divided image determined by the correction boundary value determination unit 74. The data is output to each image processing unit as use data Ua, Ub, Uc, Ud. For example, the correction data Ua for the divided image Sa includes the full-screen dark value B and the full-screen bright value W that are full-screen correction values, and unit areas b0, b5, b10, and c0 that are correction boundary values for the divided image Sa. ˜c4, d0 each dark portion attention gradation and bright portion attention gradation.

  The correction data Ua, Ub, Uc, Ud output from the correction data creation unit 104 are transferred to the image processing units 103a, 103b, 103c, 103d, respectively. The IF unit 17a of the image processing unit 103a receives the correction data Ua output from the correction data creation unit 104.

  As described above, in this embodiment, intermediate data and correction data are transferred between the plurality of image processing units and the correction data creation unit. As a bus for transferring data between a plurality of image processing units and a correction data creation unit, for example, there is a PCI (Peripheral Component Interconnect). Since the intermediate data and the correction data have a smaller data amount than a statistical amount such as a histogram, the time required for data transfer can be reduced.

  The correction data Ua received by the IF unit 17a is output to the difference value calculation unit 13a. At this time, the full screen dark value B and the full screen bright value W, which are the full screen correction values in the correction data Ua, are output not only to the difference value calculation unit 13a but also to the full screen γ generation unit 14a.

  In the γ correction in the present embodiment, the entire input image is uniformly corrected by the full screen γ curve obtained from the full screen correction value included in the correction data Ua in the full screen γ generation unit 14a. Furthermore, local correction corresponding to each pixel in the input image is performed with the local γ curve obtained from the difference value calculation unit 13a and the local γ generation unit 15a.

  First, a method for generating a full screen γ curve will be described. The full screen γ generation unit 14a generates a full screen γ curve from the full screen dark part value B and the full screen bright part value W of the full screen correction value included in the correction data Ua received by the IF unit 17a. The full-screen γ curve of this embodiment includes a dark part curve γB for correcting low gradation data, a bright part curve γW for correcting high gradation data, and an intermediate curve γM for correcting intermediate gradations.

  FIG. 9 shows the conversion characteristics of the dark part curve γB (g) and the bright part curve γW (g) which are the basis of the full screen γ curve. In this embodiment, it is assumed that the gradation resolution of the image data is 8 bits (0 to 255 gradations). The dark part curve γB (g) is a curve for γ-correcting input data having a gradation value of 0 to 128, and the bright part curve γW (g) is a curve for γ-correcting input data having a gradation value of 192 to 255. The data of the dark part curve γB (g) and the bright part curve γW (g) are stored in advance in a memory (not shown) in the image processing unit 103a.

  As shown in FIG. 9, 126 types (g: 2 to 126, 128) of dark part curves γB (g) are prepared. One dark part curve γB (B) is selected from 126 types of dark part curves γB (g) (g: 2 to 126, 128) according to the full screen dark part value B obtained by Expression 3. Since the full screen dark area value B is the minimum value of the dark area attention gradation in the representative value of each divided image as calculated by Equation 3, it takes one of values from 2 to 126 or 128. The dark part curve γB (B) is a gradation correction parameter for enhancing the gradation in the gradation value in the range around the full screen dark part value B.

  As shown in FIG. 9, 125 types (g: 128, 130 to 253) of bright portion curves γW (g) are prepared. One bright part curve γW (W) is selected from 125 kinds of bright part curves γW (g) (g: 128, 130 to 253) according to the full screen bright part value W obtained by Expression 4. Since the full-screen bright part value W is the maximum value of the bright part attention gradation in the representative value of each divided image as calculated by Expression 4, it takes one of 128, 130 to 253. The bright part curve γW (W) is a gradation correction parameter for enhancing the gradation in the gradation value in the range around the full screen bright part value W.

  By selecting the γ curve for correcting the low gradation data and the high gradation data in this way, a curve for assigning gradation to the dark part is generated as the full screen dark part value B is low, and the full screen bright part value W is high. A curve for assigning gradation to the bright part is generated as the case goes. By correcting using the γ curve determined as described above, it is possible to improve the visibility of the dark part and the bright part.

  The intermediate curve γM is a curve for γ-correcting input data with gradation values 129 to 191. The intermediate γ curve γM has a conversion characteristic obtained by interpolating the gradation between the selected dark portion curve γB (B) and the bright portion curve γW (W) by linear interpolation. As described above, in the full screen γ generation unit 14a, the full screen γ curve is generated by the dark part curve γB, the bright part curve γW, and the intermediate curve γM.

  Next, a method for generating a local γ curve will be described. The difference value calculation unit 13a receives the correction data Ua from the IF unit 17a, and receives the dark part attention gradation Nba (i) and the bright part attention gradation Nwa (in each unit area of the divided image Sa from the attention gradation detection unit 11a. i) is received. That is, the difference value calculation unit 13a includes the full-screen correction value including the full-screen dark part value B and the full-screen bright part value W, and each unit area of the divided image Sa and the other divided images in contact with the divided image Sa. The dark part attention gradation and the light part attention gradation in each of the unit areas are received. Then, using the received value, a dark part difference value that is a difference between the dark part attention gradation of each unit area and the full screen dark part value B, and a bright part attention gradation and full screen bright part value W of each unit area A bright part difference value which is a difference is obtained.

  The dark part difference value and the bright part difference value are a total of 24 unit areas a0 to a14 of the divided image Sa and unit areas b0, b5, b10, c0 to c4, and d0 in which the correction boundary value is included in the correction data Ua. It is obtained in each unit area. The unit areas b0, b5, b10, c0 to c4, and d0 in which the correction data Ua includes the correction boundary value are unit areas of other divided images in contact with the divided image Sa. In the following description, the dark part difference value and the bright part difference value obtained in each unit area of a0 to a14 are the dark part difference value dba (j) and the bright part difference value dwa (j) (j = 0 to 14). . Further, the dark part difference value and the bright part difference value obtained in the unit areas of b0, b5, b10, c0 to c4, and d0 are the dark part difference value dba (j) and the bright part difference value dwa (j) (j = 15). To 23).

First, as the dark part difference value dba (j), dark part attention gradations Nba (i), Nbb (i) in the unit areas a0 to a14 and the unit areas b0, b5, b10, c0 to c4, and d0 of the divided image Sa. Each difference between Nbc (i), Nbd (i) and the full screen dark area value B is obtained. The dark part difference values dba (0) to dba (14) in the unit areas a0 to a14 in the divided image Sa are obtained by the following equations. The dark part attention gradation Nba (i) in each unit area in the divided image Sa is a value transmitted from the attention gradation detection section 11a.
dba (j) = Nba (i) -B, j = 0 to 14, i: 0 to 14 (Formula 5)

Next, dark part difference values dba (15) to dba (17) in the unit areas b0, b5, and b10 of the divided image Sb in contact with the divided image Sa are obtained by Expression 6. The dark part difference values dba (18) to dba (22) in the unit areas c0 to c4 of the divided image Sc and the dark part difference value dba (23) in the unit area d0 of the divided image Sd are obtained by the following expressions 7 and 8, respectively. . The dark part attention gradations Nbb (i), Nbc (i), and Nbd (i) in the unit areas b0, b5, b10, c0 to c4, and d0 are included in the correction data Ua transmitted from the correction data generation unit 104. The boundary value for correction of each divided image is used.
dba (j) = Nbb (i) −B, j = 15 to 17, i: 0, 5, 10 (Equation 6)
dba (j) = Nbc (i) −B, j = 18 to 22, i: 0 to 4 (Expression 7)
dba (j) = Nbd (i) −B, j = 23, i: 0 (formula 8)
The dark part difference value dba (j) (j: 0 to 23) is 24 pieces of data. Further, since the full screen dark area value B is the minimum value of the dark area attention gradation in the entire screen from Equation 3, any dark area difference value of the dark area difference value dba (j) is always 0 or more.

The bright portion difference value dwa (j) is also obtained in the same manner as the dark portion difference value dba (j). As the bright portion difference value dwa (j), the respective differences between the bright portion attention gradations Nwa (i), Nwb (i), Nwc (i), Nwd (i) and the full screen bright portion value W in each unit region. Ask for. The bright part difference values dwa (0) to dwa (14) in the unit areas a0 to a14 in the divided image Sa are obtained by the following expression. The bright part attention gradation Nwa (i) in each unit area in the divided image Sa is a value transmitted from the attention gradation detection section 11a.
dwa (j) = W−Nwa (i), j = 0 to 14, i: 0 to 14 (Equation 9)

Next, the bright part difference values dwa (15) to dwa (17) in the unit areas b0, b5, and b10 of the divided image Sb that is in contact with the divided image Sa are obtained by Expression 10. The bright part difference values dwa (18) to dwa (22) in the unit areas c0 to c4 of the divided image Sc and the bright part difference value dwa (23) in the unit area d0 of the divided image Sd are respectively expressed by the following expressions 11 and 12. Desired. The bright part attention gradations Nwb (i), Nwc (i), and Nwd (i) in the unit areas b0, b5, b10, c0 to c4, and d0 are the correction data Ua transmitted from the correction data generation unit 104. The boundary value for correction of each divided image included in is used.
dwa (j) = W−Nwb (i), j = 15 to 17, i: 0, 5, 10 (Equation 10)
dwa (j) = W−Nwc (i), j = 18 to 22, i: 0 to 4 (Formula 11)
dwa (j) = W−Nwd (i), j = 23, i: 0 (formula 12)
The bright part difference value dwa (j) (j: 0 to 23) is 24 pieces of data. Further, since the full screen bright part value W is the maximum value of the bright part noticeable gradation in the whole screen from Equation 4, any bright part difference value of the bright part difference value dwa (j) is always 0 or more. .

  The difference value calculation unit 13a transmits the obtained dark part difference value dba (j) and bright part difference value dwa (j) to the local γ generation unit 15a. The local γ generation unit 15a generates a local γ curve for performing local correction of the divided image Sa from the dark part difference value dba (j) and the bright part difference value dwa (j) calculated by the difference value calculation unit 13a. A correction value is calculated, and a local γ curve is obtained using the local correction value. A method for calculating the local correction value in each pixel in the divided image Sa in the local γ generation unit 15a will be described. The local correction value includes a local dark part value dbaP (x, y) and a local bright part value dwaP (x, y) in each pixel in the divided image Sa.

  The local dark part value dbaP (x, y) and the local bright part value dwaP (x, y), which are local correction values in each pixel, are the dark part difference value dba (j) in each unit region calculated by the difference value calculation unit 13a. And the bright part difference value dwa (j). When the local dark portion value dbaP (x, y) and the local bright portion value dwaP (x, y) change suddenly at each unit region boundary, a discontinuous portion occurs in the image after the local γ correction. Therefore, in this embodiment, the local dark part value dbaP (x is obtained by performing bilinear interpolation according to the pixel position so that the local dark part value dbaP (x, y) and the local bright part value dwaP (x, y) change smoothly. , Y) and the local bright part value dwaP (x, y).

FIG. 10 is a diagram for explaining processing for obtaining the local dark part value dbaP (x, y) in the pixel P (x, y) by bilinear interpolation. FIG. 10 shows unit areas a0, a1, a5, and a6 in the divided area Sa. The dark part difference values dba (0), dba (1), dba (5), and dba (6) output from the difference value calculation unit 13a are respectively the central pixels P0 of the unit regions a0, a1, a5, and a6. Bilinear interpolation is performed assuming that the local dark part values dbaP (x, y) of P1, P5, and P6. The dark part difference value dbaP (x, y) is a value obtained by bilinear interpolation of the dark part difference values dba (0), dba (1), dba (5), and dba (6) according to the distances x0, x1, y0, and y1. is there. That is, the local dark part value dbaP (x, y) is obtained by the following Expression 13.
dbaP (x, y) = (dba (0) × x1 + dba (1) × x0) × y1
+ (Dba (5) × x1 + dba (6) × x0) × y0 (Formula 13)
However, x0 + x1 = 1 and y0 + y1 = 1.

The local bright part value dwaP (x, y) is also obtained for each pixel by bilinear interpolation using Expression 14 in the same manner as the local dark part value dbaP (x, y). The bright part difference values of the unit areas a0, a1, a5, and a6 are dwa (0), dwa (1), dwa (5), and dwa (6), respectively.
dwaP (x, y) = (dwa0 × x1 + dwa1 × x0) × y1
+ (Dwa5 × x1 + dwa6 × x0) × y0 (Formula 14)
However, x0 + x1 = 1 and y0 + y1 = 1.

  As described above, the local gamma generation unit 15a calculates the local dark part value dbaP (x, y) and the local bright part value dwaP (x, y) as the local correction value in each pixel P (x, y).

  The local γ generation unit 15a generates a local γ curve to be corrected for each pixel based on the calculated local correction value. The local γ curve includes a dark part curve γBp (h) for correcting low gradation data, a bright part curve γWp (h) for correcting high gradation data, and an intermediate curve γMp (h) for correcting intermediate gradations.

  FIG. 11 shows the conversion characteristics of the dark part curve γBp (h) and the bright part curve γWp (h) that are the basis of the local γ curve. Data of the dark part curve γBp (h) and the bright part curve γWp (h) are stored in advance in a memory (not shown) in the image processing unit 103a.

  As shown in FIG. 11, 127 types (h: 0 to 126) of dark part curves γBp (h) are prepared. One dark part curve γBp (dbaP (x, y)) out of 127 types of dark part curves γBp (h) (h: 0 to 126) according to the local dark part value dbaP (x, y) obtained by Expression 13 Is selected. Since the local dark part value dbaP (x, y) is obtained from the dark part difference value dba (j) in each unit region, it takes any value from 0 to 126.

  As shown in FIG. 11, 126 types (h: 0 to 125) of bright portion curves γWp (h) are prepared. One bright part curve γWp (dwaP (x, y)) from 126 kinds of bright part curves γWp (h) (h: 0 to 125) according to the local bright part value dwaP (x, y) obtained by Expression 14 ) Is selected. Since the local bright part value dwaP (x, y) is obtained from the bright part difference value dwa (j) in each unit region, it takes one of 0 to 125.

  As described above, by selecting the γ curve for correcting the low gradation data and the high gradation data, the local dark portion value dbaP (x, y) that is the difference between the dark portion attention gradation of each pixel and the full screen dark portion value B is obtained. The larger the pixel, the more the curve that compresses the gradation of the dark part is selected. On the other hand, a curve for compressing the gradation of the bright portion is selected for a pixel having a larger local bright portion value dwaP (x, y), which is the difference between the bright portion attention gradation and the W full-screen bright portion value W. The intermediate curve γMp has a conversion characteristic in which the gradation between the selected dark portion curve γBp (dbaP (x, y)) and the bright portion curve γWp (dwaP (x, y)) is connected by linear interpolation. As described above, the local γ curve is created from the dark curve γBp, the bright curve γWp, and the intermediate curve γMp.

  Using the full screen γ curve generated by the full screen γ generation unit 14a and the local γ curve generated by the local γ generation unit 15a, the γ correction processing unit 16a performs γ correction processing of the divided image Sa. In the γ correction processing in the γ correction processing unit 16a, first, the entire input image is uniformly corrected with the full screen γ curve, and then the local correction corresponding to each pixel in the input image is performed with the local γ curve. The γ correction processing unit 16a outputs the divided image Sa subjected to the γ correction processing to the image combining unit 105 as a divided output image Ta.

  The processing in the image processing unit 103a has been described above, but the image processing units 103b, 103c, and 103d have the same configuration as the image processing unit 103a, and perform similar processing.

  The image combining unit 105 outputs an output image obtained by combining the divided output images Ta, Tb, Tc, and Td, which have been subjected to the γ correction processing, output from the image processing units 103a, 103b, 103c, and 103d to a display device or the like.

  Although the present embodiment describes an example in which four image processing units are used, the number of image processing units of the present invention is not limited to four. In addition, the local γ curve has been described with the configuration obtained for each pixel, but the local dark portion value dbaP (x, y) and the local bright portion value dwaP (x, y) in each pixel are set so as not to change rapidly. Alternatively, a local γ curve may be obtained for each small block including a plurality of pixels and γ correction may be performed. In addition, the method of using the dark part attention gradation and the bright part attention gradation when obtaining the representative value and boundary value of each divided image has been described. The maximum gradation value or average value of the histogram may be used.

  As described above, according to the present invention, it is possible to secure time for performing image correction processing by reducing the data transfer time between each image processing unit and the correction data generation unit. Furthermore, according to the present invention, when local γ correction is performed using a plurality of image processing units, by transferring boundary values of some unit regions that are in contact with the divided images on which other image processing units perform correction processing, It is possible to reduce the discontinuity of the image processing for each area.

(Example 2)
In this embodiment, unlike the first embodiment, an image processing apparatus that does not use the correction data creation unit 104 of FIG. 1 will be described. In the first embodiment, each image processing unit transmits the representative value and boundary value of each divided image to the correction data creating unit 104, whereas in this embodiment, each image processing unit displays the representative value and the boundary value of each divided image. The boundary value is transmitted to another image processing unit.

  FIG. 12 is a circuit block diagram of the present embodiment. An image processing apparatus 1100 illustrated in FIG. 12 includes an image dividing unit 1102, image processing units 1103 a, 1103 b, 1103 c, and 1103 d, and an image combining unit 1105. As in the first embodiment, it is assumed that one image processing unit is mounted on one IC (Integrated Circuit). The operation of the image processing apparatus 1100 will be described below.

  The image dividing unit 1102 divides the input image into a plurality of areas and creates a plurality of divided images. This embodiment will be described using the input image S shown in FIG. The resolution of the input image S is 3840 × 2160 pixels. The image dividing unit 1102 divides the input image S into four divided images Sa, Sb, Sc, and Sd as shown in FIG. 2B, and outputs them to the image processing units 1103a, 1103b, 1103c, and 1103d, respectively. The resolution of the divided images Sa, Sb, Sc, and Sd is 1920 × 1080 pixels.

  Next, the image processing unit 1103a will be described. Since the image processing units 1103b, 1103c, and 1103d have the same configuration as the image processing unit 1103a and perform similar processing, only the image processing unit 1103a will be described in the present embodiment.

  FIG. 13 is a block diagram illustrating an internal configuration of the image processing unit 1103a. The image processing unit 103a according to the first embodiment is different from the image processing unit 103a in that the intermediate data generation unit 131a is different from the processing of the intermediate data generation unit 12a in that a full screen correction value calculation unit 132a is added. In FIG. 13, blocks having the same reference numerals as those in FIG. 2 perform the same processing as in the first embodiment, and thus will not be described in detail.

  As in the first embodiment, the target gradation detection unit 11a calculates each dark part target gradation Nba from the luminance histogram Ha (i) (i: 0 to 14) of each unit area of the divided image Sa divided as shown in FIG. (I) and bright part attention gradation Nwa (i) are detected. The attention gradation detecting unit 11a outputs the detected dark part attention gradation Nba (i) and the bright part attention gradation Nwa (i) to the intermediate data creation unit 131a and the difference value calculation unit 13a.

  In the intermediate data creation unit 131a, intermediate data Mab output from the dark part attention gradations Nba (0) to Nba (14) and the bright part attention gradations Nwa (0) to Nwa (14) to the image processing units 1103b, 1103c, and 1103d. , Mac, Mad.

  In the present embodiment, the intermediate data Mab created by the intermediate data creation unit 131a is transmitted to the image processing unit 1103b, the intermediate data Mac is transmitted to the image processing unit 1103c, and the intermediate data Mad is transmitted to the image processing unit 1103d. That is, the intermediate data Mab, Mac, and Mad are other divided image intermediate data. The intermediate data Mab, Mac, and Mad are respectively composed of representative values and boundary values of the divided image Sa. As in the first embodiment, the representative value of the divided image Sa is the minimum value mina of the dark part attention gradations Nba (0) to Nba (14) and the plurality of bright part attention gradations Nwa (0) to It consists of the maximum value maxa in Nwa (14), and is obtained by equations 1 and 2.

  The boundary values included in the intermediate data Mab, Mac, and Mad in this embodiment are different from those in the first embodiment. The boundary value included in the intermediate data Mab is composed of a dark part attention gradation Nba (i) and a bright part attention gradation Nwa (i) of a unit area in the divided image Sa in contact with the divided image Sb. Assume that the input image S is divided into unit areas a0 to a14, b0 to b14, c0 to c14, and d0 to d14 as shown in FIG. In this case, the boundary values included in the intermediate data Mab are the dark part attention gradation Nba (i) and the light part attention gradation Nwa (i) of the unit areas a4, a9, and a14.

  Similarly, the boundary values included in the intermediate data Mac are the dark part attention gradation Nba (i) and the bright part attention gradation Nwa (i) of the unit areas a10 to a14 in contact with the divided image Sc. The boundary values included in the intermediate data Mad are the dark part attention gradation Nba (i) and the bright part attention gradation Nwa (i) in the unit region a14 in contact with the divided image Sd.

  When the input image includes the divided image Sa and the divided image Sb (when the input image is divided into two and the divided image Sb is on the right of the divided image Sa), the intermediate data creation unit 131a creates the intermediate data Mab. The boundary values included in the intermediate data Mab are derived from the dark part attention gradation Nba (i) and the light part attention gradation Nwa (i) of each of the unit areas a4, a9, and a14 in the divided image Sa in contact with the divided image Sb. Become.

  The intermediate data creation unit 131a outputs the created intermediate data Mab, Mac, and Mad to the image processing units 1103b, 1103c, and 1103d through the IF unit 17a, respectively, and the representative values of the divided images Sa are corrected by the difference value calculation unit 13a and the full screen correction. It transmits to the value calculation part 132a.

  The intermediate data creation unit 131b in the image processing unit 1103b creates intermediate data Mba, Mbc, and Mbd in the same manner as the image processing unit 1103a. The intermediate data creation unit 131c in the image processing unit 1103c creates intermediate data Mca, Mcb, and Mcd, and the intermediate data creation unit 131d in the image processing unit 1103d creates intermediate data Mda, Mdb, and Mdc. Each image processing unit transmits the created intermediate data to an image processing unit that processes a divided image that is in contact with a unit area having a boundary value included in the intermediate data.

  The image processing unit 1103a receives the intermediate data Mba, Mca, and Mda from the image processing units 1103b, 1103c, and 1103d through the IF unit 17a. The received intermediate data Mba, Mca, Mda is output to the difference calculation unit 13a. The representative value of the divided image Sb, the representative value of the divided image Sc, and the representative value of the divided image Sd included in each of the intermediate data Mba, Mca, and Mda are also transmitted to the full-screen correction value calculation unit 132a.

  Through the above processing, the full-screen correction value calculation unit 132a receives the representative value of the divided image Sa, the representative value of the divided image Sb, the representative value of the divided image Sc, and the representative value of the divided image Sd from the intermediate data creation unit 131a and the IF unit 17a. Receive value. Similar to the full screen correction value calculation unit 73 of FIG. 7 in the first embodiment, the full screen correction value calculation unit 132a calculates a full screen correction value including the full screen dark part value B and the full screen bright part value W. The full screen correction value calculation unit 132a transmits the calculated full screen correction value to the full screen γ generation unit 14a and the difference value calculation unit 13a.

  Through the above processing, the difference value calculation unit 13a calculates the dark part attention gradation Nba (i) and the bright part attention gradation Nwa (i) (i: 0 to 14) in each unit area of the divided image Sa as the attention gradation detection unit 11a. Receive from. Further, the intermediate data Mba received from the IF unit 17a, the dark part attention gradation Nba (i) and the bright part attention gradation Nwa (i) in each of the unit areas in the other divided images in contact with the division image Sa. Obtained from boundary values included in Mca and Mda. Further, the full screen correction value is received from the full screen correction value calculation unit 132a. That is, the difference value calculation unit 13a, like the first embodiment, the dark part attention gradation Nba (i) and the unit area of the divided image Sa and the unit area in the other divided images that are in contact with the divided image Sa. The bright part attention gradation Nwa (i) and the full screen correction value are received.

  The full screen correction value calculation unit 132a and the full screen γ generation unit 14a generate full screen gradation correction parameters for performing gradation correction processing of the divided image Sa. Further, the difference value calculation unit 13a and the local γ generation unit 15a generate a local gradation correction parameter for correcting the gradation of each pixel of the arbitrary divided image.

  Similar to the first embodiment, the difference value calculation unit 13a performs the dark portion difference value dba (j) that is the difference between the dark portion attention gradation of each unit region and the full screen dark portion value B, and the bright portion attention gradation of each unit region. And the bright part difference value dwa (j), which is the difference between the full screen bright part value W, and the like. The difference value calculation unit 13a outputs the calculated dark part difference value dba (j) and bright part difference value dwa (j) to the local γ generation unit 15a.

  As in the first embodiment, the local γ generation unit 15a generates a local γ curve. The γ correction processing unit 16a performs γ correction processing of the divided image Sa using the full screen γ curve generated by the full screen γ generation unit 14a and the local γ curve generated by the local γ generation unit 15a, and outputs a divided output image. Ta is output to the image combining unit 1105.

  The processing in the image processing unit 1103a has been described above, but the image processing units 1103b, 1103c, and 1103d have the same configuration as the image processing unit 1103a, and the same processing is performed. The image combining unit 1105 combines the γ-corrected divided output images Ta, Tb, Tc, and Td output from the image processing units 1103a, 1103b, 1103c, and 1103d, and outputs the output image to a display device or the like.

  As described above, according to the present invention, it is possible to secure time for performing image correction processing by reducing the data transfer time. Further, when local γ correction is performed using a plurality of image processing units, the boundary values of some unit areas that are in contact with the divided images on which other image processing units perform correction processing are transferred between the plurality of image processing units. This makes it possible to reduce the discontinuity of image processing for each region.

DESCRIPTION OF SYMBOLS 102 Image division part 103a-103d Image processing part 104 Correction data creation part 105 Image connection part 11a Attention gradation detection part 12a Intermediate data creation part 13a Difference value calculation part 14a Full screen gamma generation part 15a Local gamma production part

Claims (10)

A plurality of image processing units for performing image processing on a plurality of divided images obtained by dividing an input image into a plurality of regions ;
Correction data creating means for creating correction data for each divided image for correcting the gradation of each divided image,
Each image processing unit
A histogram creation means for creating a histogram in each of the previous SL plurality of divided images,
Before Symbol histogram, the intermediate data generation means for generating intermediate data including the representative value of the gradation in each divided image, a boundary value of the gradation of a region which is in contact with the other divided image in each divided image,
Have a, a correction processing means for performing processing for correcting the gradation of each of the divided image based on the correction data created by the previous SL correction data creating means,
The correction data creating means generates correction data for each divided image for correcting the gradation of each divided image from the intermediate data in each of the plurality of divided images created by the intermediate data creating means. An image processing apparatus characterized by being created . In the histogram created by the histogram creating means, the frequency of interest is a gradation value that is equal to or greater than a predetermined threshold and has a maximum value, and the local frequency variation amount is a gradation value that is smaller than a predetermined reference. It has attention gradation detection means for detecting the tone,
The histogram creating means creates a histogram in each of the unit areas obtained by dividing the divided image into a plurality of parts,
The attention gradation detection unit detects a dark part attention gradation on a low gradation side and a bright part attention gradation on a high gradation side among the attention gradations of the histogram in each unit area of the divided image. ,
The intermediate data in each divided image created by the intermediate data creating means is a representative value of the gradation in the divided image based on the dark part attention gradation and the bright part attention gradation in each unit area of the division image, and The boundary value of the gradation in the said divided image which consists of the dark part attention gradation and the bright part attention gradation in the unit area | region which touches another division image in the said division image, The said division image is characterized by the above-mentioned. Image processing device. The attention gradation detecting means sets the minimum value on the low gradation side among the attention gradations of the histogram in each unit area of the divided image as a dark portion attention gradation, The maximum value is detected as the bright part attention gradation,
The representative value included in the intermediate data in each divided image created by the intermediate data creating means is the minimum value in the dark part attention gradation and the bright part attention gradation in each unit area of the divided image. Is the maximum value of
The correction data for each divided image created by the correction data creating means includes a full-screen correction value and a correction boundary value for each divided image,
The full-screen correction value is a full-screen dark value having a minimum gradation value and a full-screen dark value having a maximum gradation value, from among the representative values of the intermediate data in each of the plurality of divided images.
The correction data of the divided image includes the dark part attention gradation and the bright part attention gradation in a unit region in another division image that is in contact with each division image from among the boundary values of the plurality of intermediate data. The image processing apparatus according to claim 2 .   The correction processing means enhances the gradation of the range around the full-screen dark value and full-screen light value included in the full-screen correction value of the correction data created by the correction data creation means. The image processing apparatus according to claim 3, wherein a gradation correction process of the divided image is performed using a full-screen gradation correction parameter for the display.   The correction processing means includes a dark part attention gradation and a bright part attention gradation in each unit area of the divided image detected by the attention gradation detection means, and the correction data created by the correction data creation means. By using a local gradation correction parameter for correcting the gradation of each pixel of the divided image created from the full-screen correction value and the correction boundary value of each divided image included, the scale of each pixel of the divided image is calculated. The image processing apparatus according to claim 4, wherein tone correction processing is performed. In an image processing apparatus that performs image processing on a plurality of divided images obtained by dividing an input image into a plurality of regions using a plurality of image processing units,
Each image processing unit
A histogram creating means for creating a histogram of the corresponding divided image;
Intermediate data creating means for creating, from the histogram, intermediate data including a representative value of gradation in the divided image and a boundary value of gradation relating to a region in contact with another divided image in the divided image;
Transmitting means for transmitting the intermediate data to another image processing unit;
Full-screen gradation that performs gradation correction processing of the divided image from the representative value of gradation in the divided image and intermediate data of the other divided image transmitted from the transmission unit of the other image processing unit Full-screen gradation correction parameter generating means for generating correction parameters;
An image processing apparatus comprising: correction processing means for performing processing for correcting gradation of the divided image based on the full-screen gradation correction parameter. In the histogram created by the histogram creating means, the frequency of interest is a gradation value that is equal to or greater than a predetermined threshold and has a maximum value, and the local frequency variation amount is a gradation value that is smaller than a predetermined reference. It has attention gradation detection means for detecting the tone,
The histogram creating means creates a histogram in each of the unit areas obtained by dividing the divided image into a plurality of parts,
The attention gradation detection unit detects a dark part attention gradation on a low gradation side and a bright part attention gradation on a high gradation side among the attention gradations of the histogram in each unit area of the divided image. ,
The intermediate data created by the intermediate data creating means includes a representative value of gradation in the divided image based on the dark part attention gradation and bright part attention gradation in each unit area of the divided image, and the divided image. 7. The image processing apparatus according to claim 6, further comprising: a gradation boundary value in the divided image, which includes a dark part attention gradation and a light part attention gradation in a unit area in contact with another divided image. . The dark part attention gradation and the bright part attention gradation in each unit area of the divided image, the boundary value included in the intermediate data of another divided image transmitted from the other image processing unit by the transmission unit, and all Local gradation correction parameter generation means for creating a local gradation correction parameter for correcting the gradation of each pixel of the divided image from the full-screen gradation correction parameter generated by the screen gradation correction parameter generation means; Have
The image processing apparatus according to claim 7, wherein the correction processing unit corrects the gradation of the divided image based on the full-screen gradation correction parameter and the local gradation correction parameter. A plurality of image processing units that perform image processing on a plurality of divided images obtained by dividing an input image into a plurality of regions, and correction data for each divided image for correcting the gradation of each divided image A correction data creating means for creating an image processing apparatus comprising:
Each image processing unit
A histogram generation step of generating a histogram in each of the previous SL plurality of divided images,
Before Symbol histogram, the intermediate data generation step of generating intermediate data including a representative value of the gradation in each divided image, a boundary value of the gradation of a region which is in contact with the other divided image in each divided image,
Have a, a correction processing step for performing a process of correcting the tone of each divided image based on the correction data created by the previous SL correction data creating means,
The correction data creation means generates correction data for each divided image for correcting the gradation of each divided image from the intermediate data in each of the plurality of divided images created in the intermediate data creation step. method of controlling an image processing apparatus characterized by have a step of creating. In a control method for an image processing apparatus that performs image processing on a plurality of divided images obtained by dividing an input image into a plurality of regions using a plurality of image processing units,
Each image processing unit
A histogram creation step for creating a histogram of the corresponding divided image;
An intermediate data creating step for creating intermediate data including a representative value of gradation in the divided image and a boundary value of gradation relating to a region in contact with another divided image in the divided image from the histogram;
A transmission step of transmitting the intermediate data to another image processing unit;
A full-screen gradation correction parameter for performing gradation correction processing of the divided image from a representative value of gradation in the divided image and intermediate data of the other divided image transmitted from the other image processing unit. A full-screen gradation correction parameter generation step to be created;
And a correction processing step for performing a process of correcting the gradation of the divided image based on the full-screen gradation correction parameter.

Images