JP4428895B2 - Image processing apparatus, method, program, and recording medium - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an image processing apparatus, and more particularly to an image processing apparatus that classifies a given image into any of various types of images (character originals, photographs, diagrams, designs, etc.) and converts them into density characteristics suitable for each image. .
[0002]
[Prior art]
In general, as a processing function of an image input device or a manuscript copying device, input image data is classified by image type, and filtering processing and gamma correction correction (histogram optimization) suitable for the properties of the manuscript image to be processed are performed. It has been practiced to improve image quality.
[0003]
Filter processing emphasizes or improves the local characteristics of the image to improve the image quality of the original image. An integral filter that reduces noise and a differential filter that enhances the edges in the image Etc.
[0004]
The gamma correction mentioned as another process is a process for correcting gradation of image density. Examples of density gradation correction include processing for uniformly expanding or contracting the range of density values, nonlinear conversion for expanding or compressing a specific density region from other portions, and the like.
[0005]
Various processing methods for classifying image types have been devised. In many of these cases, a feature amount presented by the type of image is extracted, and the feature amount is determined by a predetermined evaluation function or determination function to determine the image type. Conventionally, there are many examples in which an image includes black pixels in a predetermined block region, edge occurrence frequency, density level histogram, spatial frequency distribution, line segment direction distribution, and the like as feature amounts.
[0006]
Also in the present invention, Japanese Patent Publication No. 4-18350 discloses an example in which the frequency distribution of the density gradient of the input image used as the feature quantity is used as the feature quantity. In this classification processing method, a density gradient is obtained for each pixel unit in the horizontal and vertical directions of a digital image that is an input image, and the direction calculated from the obtained horizontal and vertical density gradient values is counted within the region. The frequency distribution is obtained. The variance of the frequency is calculated from the frequency distribution, and whether or not the character area is determined is determined by threshold determination between the variance and a predetermined threshold.
[0007]
[Problems to be solved by the invention]
In order for the above-described conventional method to classify image data satisfactorily, the input image data is intended for ideal image data in a relatively good state containing no noise. In practice, however, illumination irregularities and document stains often enter the input image data, resulting in partial degradation of contrast and noise in such images, making black pixel detection and edge extraction unstable. Become. Therefore, in the conventional method using unstable black pixel detection and edge extraction as parameters, it is very difficult to accurately determine the image type.
[0008]
Further, even if the threshold determination is performed only on the dispersion with respect to the direction distribution of the density gradient, for example, the brightness (density) level range is narrow, so-called low-contrast images, or the edge of the character with respect to the determination target small area In the case where the proportion of the portion is small or the thickness of the line of the character itself is thin, even if it is a printed character, the dispersion becomes small, and a clear determination becomes difficult. This is because the direction distribution of the density gradient exhibited by the background increases as a frequency, and a relative difference from that exhibited by the character portion is not recognized. Since the direction distribution of the density gradient indicated by the background has no normal direction dependency, the direction dependency of the direction distribution of the edge of the character portion is buried in the background distribution.
[0009]
Furthermore, not only typeface character images but also various types of images (handwritten characters, photographs, designs, backgrounds) are selected and classified, it is not possible to distinguish them at all simply by observing the distribution of the direction distribution of the density gradient. there were.
[0010]
In addition, since the document type is discriminated using the local feature amount of each original image, even if the color / monochrome identification is performed on the preview image, the text original and the photo original are identified with a high resolution main scan. Had to do with images. Alternatively, the preview must be read at the same resolution as the main scan.
[0011]
The present invention has been made in view of such points, and its purpose is to determine the type of document based on the image feature amount of the entire specified range on the input document image, and in particular, a preview image having a relatively low resolution. Therefore, it is an object of the present invention to provide an image processing apparatus or an image processing method, and a storage medium that can reliably classify character images and photographic images having different density processes and do not cause a large failure even in erroneous determination.
[0012]
[Means for Solving the Problems]
In order to achieve the above object, the present invention has the following configuration.
[0013]
An image processing apparatus that determines a type of an input image and performs processing according to the determined type of image, a detection unit that detects a density difference between adjacent pixels of the input image , and all of the input image calculating means for calculating a frequency distribution of the values of the density difference detected for the pixels, and calculating means for calculating a first-order function approximating the frequency distribution, characteristic amount detection means for detecting an inclination of the linear function as a feature And an image processing apparatus comprising: a determination unit configured to determine the image content of the input image as one of a plurality of image types by comparing the feature amount with a predetermined threshold value .
[0021]
An image processing method for determining a type of an input image and performing a process according to the determined type of image , wherein a detection step for detecting a density difference between adjacent pixels of the input image and all of the input images a calculation step of calculating a frequency distribution of the values of the density difference detected for the pixel, the calculation step of calculating a first-order function approximating the frequency distribution, the feature quantity detecting step of detecting a tilt of the linear function as a feature And a determination step of determining the image content of the input image as one of a plurality of image types by comparing the feature amount with a predetermined threshold value .
[0029]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments to which the present invention is applied will be described in detail with reference to the drawings.
[0030]
(First embodiment)
First, an outline of the operation of the image processing apparatus according to the first embodiment will be described with reference to a flowchart, and then a document type determination process which is a main requirement of the present invention will be described in detail.
[0031]
The image processing apparatus according to the present embodiment determines that the type of document image is one of {text document, photo document, other document} based on the preview image read at a low resolution, and the density prepared for each document type. It has a function of calculating a conversion table (hereinafter referred to as a γ table) and optimizing the density of the main scan image that is subsequently read.
[0032]
FIG. 1 is a flowchart showing a schematic processing procedure of the image processing apparatus according to the first embodiment.
[0033]
The image processing apparatus inputs a relatively low resolution preview image read by an image reading apparatus or the like, and simultaneously receives the position of the main scan reading range to be finally stored on the preview image (step S101).
[0034]
The image processing apparatus determines that the image type of the reading designated range on the preview image is one of {text original, photo original, other original} by an image type determination procedure described in detail later in this specification ( Step S102). When all the pixels in the specified range of the preview image have the same value, for example, when it is pure white or black, it is determined that the judgment is impossible and the density distribution of the original is not adjusted, and a so-called through γ table is created. (Steps S103 and S105). In the case of a through γ table, the input value is directly used as the output value.
[0035]
When the determination of the image type is completed normally (step S104), the γ table is calculated according to the result and the density distribution within the designated range of the preview image (steps S106, S107, S108). Details of this will be described later.
[0036]
Next, the density conversion of the preview image is performed using the γ table for optimizing the density distribution for each original calculated based on the density distribution of the preview image (step S109), and the correction result is displayed to the operator (step S109). Step S110). If it matches the operator's intention, the process proceeds to a request for the actual scan data. If necessary, the preview image display is updated again after fine adjustment of the γ table (step S112) (step S110) and an instruction is waited for (step S110). S111).
[0037]
Subsequently, the image reading apparatus reads the original with the resolution desired by the operator (step S110, main scan operation). The image processing apparatus that has received the original image (the main scan image) uses the previously determined γ table to optimize the density distribution, thereby performing the density gradation adjustment optimal for the content of the original image. Data is created (step S114).
[0038]
In the above description, the image processing apparatus optimizes the density distribution of the main scan image using the γ table. However, if the image reading apparatus has a function to accept and process the γ table, the main scanning operation is performed prior to the main scanning operation. A configuration in which the γ table is transferred from the image processing apparatus to the image reading apparatus and the density distribution is optimized by the image reading apparatus is also possible.
[0039]
Next, an image type determination procedure in the present embodiment will be described in detail with reference to a flowchart and a frequency distribution graph.
[0040]
FIG. 2 is a flowchart showing document type determination processing of the image processing apparatus according to the first embodiment.
[0041]
FIGS. 6 and 7 show the frequency of density difference between the R color density frequency distribution (0 is black and 255 gives white) of the photographic image and the text image and the adjacent pixel in the X direction of the R color (the horizontal direction of the image). The distribution is displayed. The horizontal axis indicates the density of each pixel for the density frequency distribution displayed by a thin line, and the absolute value of the density difference for the frequency distribution of the density difference displayed by a thick line. However, the vertical axis | shaft of FIG. 6, FIG. 7 displays the ratio of the frequency with respect to the total pixel number by the logarithm which makes 10 the base. That is, pixels having a density value of −3 on the vertical axis occupy a ratio of 0.1% of all pixels.
[0042]
FIG. 8 illustrates a procedure for calculating the feature value K _RX from the frequency distribution of the density difference between adjacent pixels in the X direction of the R color in an example of the frequency distribution of an image in which photographs and characters are mixed.
[0043]
As shown in the flowchart of FIG. 2, in the image type determination process, while calculating the frequency distribution of density for each color, between adjacent pixels in the vertical direction (Y-axis direction) or the horizontal direction (X-axis direction). The density difference is calculated and a frequency distribution of density differences is created (step S201).
[0044]
After the calculation of the density difference frequency distribution for all pixels, a process of doubling the frequency of the density difference 0 is performed. This is intended to correct the probability that the density difference becomes zero while the density difference of 1 or more includes both positive and negative density changes in order to calculate the density difference as an absolute value. .
[0045]
The density frequency distribution (taken as an example of the density frequency distribution of the R color) and the density difference frequency distribution (taken as an example of the density difference frequency distribution of the R color in the X direction) created by the above procedure are shown in FIG. 7 or the distribution shown in FIG. In the photographic document shown in FIG. 6, the density difference frequency monotonously decreases as the density difference increases, whereas in the text document shown in FIG. 7, there is a range in which almost the same density difference frequency exists regardless of the density difference. Yes (almost horizontal part of the thick line graph).
[0046]
In this document type discrimination process, paying attention to the characteristics in the density difference frequency distribution, the central part of the density difference frequency distribution is approximated by a linear function, and the inclination is adopted as the feature amount parameter. Hereinafter, the calculation procedure will be described with reference to FIG.
[0047]
The minimum value of the density difference frequency distribution (about -6 in FIG. 8), the graph indicating the density difference frequency distribution, and the total accumulated value of the portion surrounded by the Y axis are calculated. Next, from the density difference 0 side, a density difference in which the cumulative value becomes 10%, 20%, 40%, 60%, and 80% of the total cumulative value is searched, and the interval [10% to 20%] and the interval [20% ˜40%], section [40% -60%], and average value in section [60% -80%] are calculated to determine four representative points indicated by white ellipses in FIG. Then, a straight line f _RX (d) = K _RX * d + B _RX (4) passing through four representative points
Is calculated by the method of least squares.
[0048]
Similarly, a linear function representing the characteristics of each density difference frequency distribution is calculated using the density difference frequency distribution in the Y direction, or the density difference frequency distribution of G and B colors, and the slope coefficient {K _RX , K _GX , K _BX , K _RY , K _GY , K _BY } are obtained (step S202).
[0049]
Using the feature amount parameter calculated in the above procedure, the document type is determined in accordance with a plurality of given determination criteria (step S203). Next, an example of the criterion is given.
[0050]
Criterion 1:
If {K _RX <T _p1 } and {K _GX <T _p1 } and {K _BX <T _p1 }
And {K _RY <T _p1 } and {K _GY <T _p1 } and {K _BY <T _p1 }
Then, criteria 2 for a photo manuscript:
If {K _RX > Tt ₁ } and {K _GX > Tt ₁ } and {K _BX > Tt ₁ }
And {K _RY > Tt ₁ } and {K _GY > Tt ₁ } and {K _BY > Tt ₁ }
Then, criterion 3 which is a text manuscript:
If {{{K _RX > Tt ₁ } and {K _RY > Tt ₁ }}
Or {{K _GX > Tt ₁ } and {K _GY > Tt ₁ }}
Or {{K _BX > Tt ₁ } and {K _BY > Tt ₁ }}}
And {{K _RX > T _p2 } and {K _GX > T _p2 } and {K _BX > T _p2 }
And {K _RY > T _p2 } and {K _GY > T _p2 } and {K _BY > T _p2 }
If it does not meet any of the determination criteria for a text document, the document is determined as another document.
[0051]
However, T _p1 , T _p2 and Tt ₁ are given constants experimentally determined in advance.
[0052]
The determination criterion 1 and the determination criterion 2 indicate conditions under which all of the feature amount parameters in each color and each direction are a photographic document or a text document. Further, in the criterion 3, it is determined that any feature parameter is not a photograph, and a composite condition is given in which any one color is determined to be a text document. This is a criterion for detecting a document in which characters are displayed in only one color, and it is possible to further improve the accuracy of the determination by adding a criterion according to the characteristics of the existing document. is there.
[0053]
Next, a procedure for creating a density conversion table optimal for the document density distribution when the document type is determined will be described with reference to a flowchart.
[0054]
FIG. 3 is a diagram for explaining the creation of a density conversion table (γ table) suitable for a text image according to the first embodiment.
[0055]
First, the frequency distribution of density is calculated independently for each color regardless of the color. Here, the density frequency distribution for each color previously calculated in the flowchart (FIG. 2) representing the document type determination process can be easily obtained by adding (step S301).
[0056]
Next, in the density frequency distribution synthesized from the density frequency distribution of each color (hereinafter simply referred to as the density frequency distribution), the minimum value (hereinafter referred to as the Shadow value) and the maximum value (hereinafter referred to as the Shadow value) where the frequency distribution exists. Is called a Highlight value), and a value that bisects the frequency distribution (hereinafter called the Median value) is searched (step S302).
[0057]
The γ table is created as follows. The value in the range below the Shadow value is 0, the value in the range exceeding the Highlight value is 255, the range from the Shadow value to the Median value is convex downward, and the range from the Median value to the Highlight value is an upward convex curve Created as When the frequency distribution is relatively uniformly distributed from the Shadow value to the Highlight value, the degree of convexity is set strongly, and a γ table as shown in FIG. 9 is calculated (step S303). With this γ table, the black side is converted darker than the median value, and the white side is converted brighter than the median value.
[0058]
On the other hand, in the frequency distribution from the Shadow value to the Highlight value, if the frequency near the Median value is small and the frequency near the Shadow value or the Highlight value is large, there is sufficient contrast in the original image. Therefore, the relatively convex γ table as shown in FIG. 10 is calculated, and the gradation characteristics of the original image remain.
[0059]
According to the algorithm for creating the above-described density conversion table, it is possible to obtain a density conversion table with high contrast and high visibility for a text image.
[0060]
Also, since the density frequency distribution of the original image is reflected in the S-shape γ, the original document with a small intermediate density frequency distribution has a small amount of density adjustment, whereas the original document with a large intermediate density frequency distribution has a stronger contrast adjustment. Is done.
[0061]
FIG. 4 is a diagram for explaining creation of a density conversion table suitable for a photographic image according to the first embodiment, and FIG. 11 is a diagram showing an outline of calculation contents.
[0062]
First, the frequency distribution of density is calculated independently for each color regardless of the color. Here, it is possible to easily obtain the density frequency distribution for each color calculated previously in the flowchart (FIG. 2) representing the document type discrimination process (step S401).
[0063]
Next, in the density frequency distribution synthesized from the density frequency distribution of each color (hereinafter simply referred to as the density frequency distribution), the minimum value (hereinafter referred to as the Shadow value) and the maximum density value where the frequency distribution exists from the 0 side. While searching for a value (hereinafter referred to as “Highlight value”) (step S402), the cumulative value of the frequency is calculated (step S403).
[0064]
The γ table is created as follows. The value in the range below the Shadow value is 0, and the value in the range above the Highlight value is 255. Further, the range from the Shadow value to the Highlight value is expressed by the function g (d) = (d Shadow) ^ G (Expression 2)
Thus, a multiplier G that can be approximated most appropriately is obtained by the method of least squares (step S404).
[0065]
The operator A ^ B is an operator representing A to the Bth power.
[0066]
G ⁻¹ (d) = (dShadow) ^ (1 / G) (Formula 3) which is an inverse function of Formula 2
Can be used to obtain a power function that uniformly converts the density distribution, but the tone of the original image is also lost, so g ′ (d) = (dShadow) ^ using a given parameter T ((1 / G) ^ T), 0 <T <1.0 (Formula 4)
Calculate with. T is a given parameter that controls the degree of density correction, and changes the density distribution to be the most uniform when T = 1.0, and does not affect the density distribution at T = 0.0 ( Step S405).
[0067]
The calculation results obtained in this way are γ tables as shown in FIGS. For example, if the density distribution is extremely biased to the black side, the γ table is as shown in FIG. 12, and if the density distribution is slightly biased to the white side, the γ table is as shown in FIG.
[0068]
According to the above-described algorithm for creating a density conversion table, a density conversion table for increasing the contrast of density at which the frequency distribution is concentrated can be obtained for a photographic image, so that it is possible to convert the photographic image to a well-graded photographic image. .
[0069]
As described above, according to the image processing apparatus of the present embodiment, after calculating the frequency distribution individually for each color by calculating the density difference with adjacent pixels in the specified range on the input document image in the vertical direction, A feature quantity using a coefficient of a linear function obtained by calculating a linear function approximating each logarithmically transformed frequency distribution as a parameter is synthesized to determine one of the image types or an undecidable image. As a result, it is possible to create a density conversion table optimal for the image and optimize the density gradation.
[0070]
(Second Embodiment)
FIG. 5 is a flowchart showing a schematic processing procedure of the image processing apparatus according to the second embodiment.
[0071]
The image processing apparatus inputs a main scan image read at a resolution intended for storage or printing by an image reading apparatus or the like (step S501).
[0072]
The image processing apparatus temporarily stores the input image in a data storage device such as a memory or a magnetic storage device, and performs an image of the designated reading range on the input image by the image type determination procedure described in detail in the first embodiment. The type is determined to be any one of {text original, photo original, other original} (step S502). When all the pixels in the specified range of the input image have the same value, for example, when the pixel is completely white or completely black (step S503), it is determined that the determination is impossible and the density distribution of the document is not adjusted, so-called through γ A table is created (step S505). In the case of a through γ table, the input value is directly used as the output value.
[0073]
When the image type determination is completed normally, the γ table is calculated according to the result and the density distribution within the designated range of the preview image (step S504) (steps S506, S507, and S508).
[0074]
Subsequently, the image processing apparatus reads out the input image temporarily stored in the data storage device and optimizes the density distribution by using the previously determined γ table, so that the density density optimum for the content of the original image is obtained. Image data that has been adjusted is created (step S509).
[0075]
According to the second embodiment, in an image reading apparatus that does not necessarily input a preview image, for example, an ADF (automatic document feeder) or a color FAX, a document image is continuously adjusted while adjusting the balance of the density of the entire image. Can be processed.
[0076]
In the image processing apparatus and the image processing procedure described in the first and second embodiments, the processing is performed on the assumption that the designated range is the same image type. However, prior to performing the determination according to the present invention, it is simple. It is within the scope of the present invention to perform the determination according to the present invention for each part after the image is divided into similar parts by an evaluation function or determination function.
[0077]
The procedure for dividing the image features into similar parts by the simple evaluation function or determination function mentioned above is a technique conventionally known as image area separation processing, and will not be described in detail here.
[0078]
In order to realize the functions of the image processing apparatus and the image processing procedure described in the first and second embodiments, various components of the image processing apparatus are operated and each step of the image processing procedure is executed. The program code itself and means for supplying the program code to the computer, for example, a storage medium storing the program code belong to the category of the present invention.
[0079]
In this case, the program code stored in the storage medium is read out by a predetermined storage / reproduction device, and the EEPROM operates. As a storage medium for storing the program code, for example, a floppy disk, a hard disk, an optical disk, a magneto-optical disk, a CD-ROM, a magnetic tape, a nonvolatile memory card, a ROM, or the like can be used.
[0080]
Further, by executing the program code supplied by the computer, not only the functions of the embodiments are realized, but also the OS (operating system) or other application software running on the computer. Such a program code is also included in the present invention even when the functions of the respective embodiments are realized together.
[0081]
Further, after the supplied program code is stored in the memory provided in the function expansion board of the computer or the function expansion unit connected to the computer, the CPU provided in the function expansion board or function expansion unit based on the instruction of the program code A system in which the functions of the embodiments are realized by performing part or all of the actual processing and the like is also included in the present invention.
[0082]
【The invention's effect】
According to the present invention, a document type is discriminated based on the feature amount of the designated entire range of input image data, and even when a preview image having a relatively low resolution is used, a character image and a photograph with different density processing are used. It is possible to classify images reliably and to optimize the image density without causing a major failure even when erroneous determination is made.
[Brief description of the drawings]
FIG. 1 is a flowchart showing a schematic processing procedure of an image processing apparatus according to a first embodiment of the present invention.
FIG. 2 is a flowchart showing document type determination processing of the image processing apparatus according to the first embodiment of the present invention.
FIG. 3 is a diagram for explaining creation of a density conversion table suitable for a text image according to the first embodiment of the present invention.
FIG. 4 is a diagram illustrating the creation of a density conversion table suitable for a photographic image according to the first embodiment of the present invention.
FIG. 5 is a flowchart showing a schematic processing procedure of the image processing apparatus according to the second embodiment of the present invention.
FIG. 6 is a graph showing a density frequency distribution of R color of a photographic image and a frequency distribution of density differences between adjacent pixels of the R color in the X direction.
FIG. 7 is a graph showing a density frequency distribution of an R color of a text image and a frequency distribution of a density difference between adjacent pixels of the R color in the X direction.
FIG. 8 is a diagram illustrating data processing in an image type determination procedure in the embodiment of the present invention.
FIG. 9 is an example of a γ table suitable for a text document obtained by an embodiment of the present invention.
FIG. 10 is another example of a γ table suitable for a text document obtained by an embodiment of the present invention.
FIG. 11 is a diagram illustrating data processing in a procedure for calculating a γ table suitable for a photographic image in the embodiment of the present invention.
FIG. 12 is an example of a γ table suitable for a photographic document obtained by an embodiment of the present invention.
FIG. 13 is another example of a γ table suitable for a photographic document obtained by an embodiment of the present invention.

Claims (9)

Determines the type of the input image, an image processing apparatus for performing a process in accordance with the type of the determined image,
Detecting means for detecting a density difference between adjacent pixels of the pixel of the input image ;
Calculating means for calculating a frequency distribution of the values of the density difference detected for all the pixels of the input image,
Calculating means for calculating a first-order function approximating the frequency distribution,
Feature quantity detection means for detecting the slope of the linear function as a feature quantity;
An image processing apparatus comprising: a determination unit that determines the image content of the input image as one of a plurality of image types by comparing the feature amount with a predetermined threshold value . The determination means determines that the input image is an image of a photographic document when the feature amount is smaller than a first threshold value, and the input image when the feature amount is larger than a second threshold value. The image processing apparatus according to claim 1, wherein the image processing apparatus determines that the image is an image of a text document . 2. The image processing apparatus according to claim 1 , further comprising means for calculating a plurality of density conversion algorithms, wherein density conversion corresponding to the image type determined by the determination means is performed on image data obtained by reading the original of the input image. The image processing apparatus described. Resolution of the image using the input image to the one determined the plurality of image type is according to claim 1 which comprises using a lower resolution than the resolution of the processing performed image corresponding to the determination Image processing device. The input image is an image in a range designated on a preview image as a range for performing the main scan reading, and the image to be processed according to the determination is an image obtained by the main scan reading. The image processing apparatus according to claim 4. The image processing apparatus according to claim 1, wherein the input image is an image in a specified range among images obtained by the main scan reading. A program for causing a computer to function as a component of the image processing apparatus according to claim 1 .   A recording medium on which the program according to claim 7 is recorded. Determines the type of the input image, an image processing method for performing a process in accordance with the type of the determined image,
A detection step of detecting a density difference between adjacent pixels of the pixels of the input image ;
A calculation step of calculating a frequency distribution of the values of the density difference detected for all the pixels of the input image,
A calculation step of calculating a linear function that approximates the frequency distribution,
A feature amount detection step of detecting the slope of the linear function as a feature amount;
An image processing method comprising: a determination step of determining the image content of the input image as one of a plurality of image types by comparing the feature amount with a predetermined threshold value .

Images