JP2940913B2 - Color image data encoding device - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to an apparatus for encoding color image data. [Prior Art] Conventionally, when transmitting or storing an image, it is common to suppress the redundancy by encoding in consideration of the efficiency. Not only black and white and binary images, but also color images and the like have an enormous amount of information, and their encoding is essential.
Particularly, in the case of a color image, since the amount of information is large, it is desired to realize an efficient encoding method. [Problems to be Solved by the Invention] However, a method of efficiently encoding color image data in which each color component data is represented by a multi-value has not been established yet. The present invention has been made to solve the above problems, and an object thereof is to achieve both efficient encoding of a plurality of components of a color image and maintenance of the image quality of the encoded image. [Means for Solving the Problems] In order to achieve the above object, a color image data encoding apparatus according to the present invention comprises: a division unit for dividing color image data into blocks; First extraction means for extracting information and color information related to a color; and extracting the average brightness information in the block and the structure information of the brightness in the block by orthogonally transforming the brightness information. A second extraction unit for extracting average color information in the block from the color information; and a quantization unit for compressing an information amount by quantization, wherein the extracted average brightness information and structure information are extracted. And quantization means for quantizing the average color information and the average color information in a setting of a quantization rate for the average brightness information> a quantization rate for the average color information> a quantization rate for the structure information of the brightness. And butterflies. Embodiment An embodiment according to the present invention will be described below in detail with reference to the accompanying drawings. <Principle of the embodiment> Since color information in the RGB color system has a strong correlation between RGB signals, there is a property that a large amount of information is lost when R, G, and B are individually compression-coded. Therefore, the RGB signal system is not suitable for compression coding. Therefore, in the present embodiment, the RGB primary color system color image data is converted into a color system having less correlation between signals, and this new color system color image data is cut out into small blocks. Then, as shown in FIG. 2, compression encoding is performed for each block.
In FIG. 2, L represents lightness information on the lightness in the block,
S represents structural information on a structure such as an edge in the block, and C represents color information on a color in the block. As described above, by extracting and encoding brightness information, structure information, and color information from color image data of the RGB color system, highly efficient compression encoding is achieved. Further, when the color image data encoded as described above is converted, for example, when only the brightness is converted, only the brightness information portion needs to be converted as shown in FIG. In this case, only the color information portion needs to be converted as shown in FIG. 4, and the above-mentioned conventional disadvantage is solved. By the way, in the embodiment described below, the L ^* a ^* b ^* color system of the CIE1976 uniform color space is adopted as the color system with little correlation between signals. <RGB → Conversion to L ^* a ^* b ^* > FIGS. 5 (a) and 5 (b) show the RGB system in the target image.
The conversion to the L ^* a ^* b ^* system and the cutting out of 4 × 4 blocks are shown. 201 is a manuscript, "A" is in manuscript 201
Is drawn. Reference numeral 202 denotes a block which is cut out in a 4 × 4 size in order from the corner of the document. 203 is
This is one of the blocks on the character "A", and indicates a case where an ending part is included. FIG. 5B shows the distribution of color image data for each RGB of the block 203, and particularly shows a case where the character of the block 203 is a red character. In this case, the RGB primary colors are as shown in the figure, and an edge appears only in R. FIG. 5C shows a case where the RGB signals shown in FIG. 5B are converted into L ^* a ^* b ^* . In the figure, L ₀ ~L _F, etc. are shown, the L ^* component in block. The letters _A to _F in the suffix represent 10 to 15 for convenience. Here, a conversion formula for converting RGB into an L ^* a ^* b ^* signal is shown below. _{X = X r R + X g} G + X b B Y = Y r R + Y g G + Y b B Z = Z r R + Z g G + Z b B _{where, X r, X g, X} b, Y r, Y g, Y b, Z b , etc. Is a constant. From this, L ^* = 116 · (Y / Y ₀ ) ^1/3 −16 a ^* = 500 · (X / X ₀ ) ^1/3 − (Y / Y ₀ ) ^1/3 b ^* = 200 · (Y / Y ₀ ) ^1/3 − (Z / Z ₀ ) ^1/3 where X ₀ , Y ₀ , and Z ₀ are values of reference white light, and Y / Y ₀ > 0.00
8856. ^{<L *} compression encoding> ^L * ^a * ^{b *} system said that less inter-signal correlation. From the L ^* a ^* b ^* system, an orthogonal transform, particularly a Hadamard transform and a discrete COS transform are suitable for encoding as shown in FIG. That is, since L ^* includes structural information in addition to brightness information, if the above orthogonal transformation is performed, a brightness information component and a structural information component are extracted from L ^* . In the following embodiment, a quadratic Hadamard transform is used among these orthogonal transforms. A general second-order Hadamard transform is represented by {F} = (1 / m · n) ^1/2 {H} L {H ^T }. Here, {L}: an original matrix of m × n, {H}: Hadamard matrix, {H ^T }: transposed matrix of {H}, and a transformed matrix of {F}: m × n. here, As, when the {L} aforementioned ^L * ^a * ^{b *} system of ^{L *} a, {F} was obtained by Hadamard converting from {L}, brightness information, the image data extracted structure information table Will be.
In this case, (1 / mn) _1/2 = 4. Also, for convenience,
If {L｝ ｛F} is expressed in vector, the above equation becomes However, i = 0 to 15 _(i.e., 0 _{~ F)} in, H _ij is 16 × 16Had
Represents an amard matrix. Therefore, the above equation is Becomes In the Hadamard matrix, + represents 1 and-represents-.
Represents 1. As can be seen from the above equation, F ₀ is the average lightness in the block,
That is, it represents the brightness information of the block. Also, F other than F _0
i (i = 1 to F) represents structural information such as an edge included in the block. <Embodiment of Encoding Circuit> FIG. 1 shows an embodiment of encoding according to the present invention.
Reference numeral 301 denotes a 4-line buffer for temporarily storing an input RGB signal for block extraction. In other words, block signals are extracted by reading out the stored signals for four lines in a 4 × 4 size. 302 is RGB → L ^* a ^*
b ^{* A} circuit that performs conversion, and is converted based on the above-described conversion formula. A specific example is shown in the circuit of FIG. 6, and the conversion can be realized by the look-up table method of 401, 402, 403 in FIG. 303 is an L ^* output from the L ^* a ^* b ^* conversion unit 302, block L ₀ of L ^* as shown in FIG. 5 _{(c), L 1, ...} , a signal is output in the order of L _F is there. In FIG. 1, reference numeral 304 denotes an orthogonal transform unit, which has Had as an orthogonal transform.
FIG. 7 shows a specific circuit for performing amard conversion. In Figure 7, 410 is a Hadamard matrix address generator for generating a row address when performing matrix operation, in order to perform the matrix operation, the input L _i
Hij is output in synchronization with. Hadamard to the input L _i is 411, etc.
This is a lookup table (LUT) that multiplies the matrix coefficients and outputs the result.
When the amard matrix address generator 410 addresses the look-up table in synchronization with the input L _i , the look-up table outputs Hadamard coefficients and outputs L _i
And Hadamard coefficients are multiplied. An adder 412 performs cumulative addition. For example, in the adder 411, L ₀ + L ₁ + L ₂ +
... + L _F is calculated. 413 is a 1/4 divider. Below, 41
5 to 418 are the same, and there are a total of 16 pairs. That is, present in each F _i, following such operations are performed. In FIG. 1, 305 denotes the F ₀ above, this is a factor that represents the brightness of blocks, as described above. 307 is a quantizer for quantizing which quantizes the 10 bits of the F ₀ to 8 bits, and outputs the L (lightness information "308. This quantization, brightness information of the maximum 2 number ¹⁰ states , is compressed to a maximum 2 number ^eight states. hereinafter, in this specification, referred to as the ratio of the number of possible states of the data before and after the quantization with the quantization index. 306 F ₀ non F ₁ to F _F This is a coefficient representing the structure of the edge included in the block, as described above, and is coded into 12 bits by the quantizer 309. That is, the coefficient is predetermined as the structure information (S). In Fig. 1, reference numerals 311 and 312 denote averaging circuits for calculating block averages of a ^* and b ^* output from the L ^* a ^* b ^* converter 302, respectively. 313 is a
^This is a quantizer that quantizes the block average value of ^* b ^* collectively, and quantizes it to 12-bit color information (C). In addition, 30
It is known that the efficiency of any of the quantizers 7,309,313 is generally higher if it is constituted by a vector quantizer. 315 of FIG. 1 is L (lightness) 308, S
(Structure) 310, a multiplexer that combines C (color information) 314 into one code. 316 is the output signal of this, that is,
This is the encoded code shown in FIG. This is the transmission path,
Or it will be sent to a storage. The color image data coded in this way has the same advantages as the data compressed in a highly efficient manner, as well as the data conversion described below. . <Data Conversion = Decoding> As described with reference to FIGS. 3 and 4, color image data encoded as shown in FIG. 2 is convenient for brightness conversion or color conversion. First, the brightness conversion will be described. <Brightness Conversion> In FIG. 8, reference numeral 320 denotes a transmission line or a storage, and 321 denotes a DMUX which performs the reverse operation of the multiplexer 315. That is, serial brightness information, structure information, and the like are converted into parallel. Reference numeral 322 denotes a conversion unit for performing brightness conversion, and 323 to be input to the brightness conversion unit 322.
Is used to input the degree of brightness conversion, and is a control amount of 8 bits. FIG. 9 shows an example of a circuit in the case where the brightness conversion section 322 is configured by a look-up table (LUT). When this lightness conversion is performed, L changes to L '. FIG. 10 shows an example of an 8-bit lightness conversion amount 323. In this case, the range of possible values is 0 to 255. The total value can be changed to 256 levels, and 128 is the default value without change. Reference numeral 325 in FIG. 8 denotes an orthogonal inverse transformer, which can be configured with the same hardware as the orthogonal transformer 304 in FIG. 1 used at the time of encoding. 326 is for decoding S (structure information), 327 is its output, 328 and 329 are for a ^* b ^* decoding, 330,331 and 332 are respectively decoded L ^* a ^* b ^* signals, and 333,335 and 337.
Are the output lines of the decoded R ', G', B ', respectively, for decoding R', G ', B'. In this way, brightness conversion is easily performed. <Color Conversion> FIG. 11 shows the data format of C (color information).
C (color information) is represented by hue (θ) and saturation (h) in a ^* -b ^* space. The relationship between hue h and saturation θ is
This is as shown in FIG. FIG. 12 (c) shows a ^* -b
^* The state of division in space is shown, and each grid point as indicated by 501 is selected as a representative color by the quantization circuit 313 in FIG. FIG. 13 shows an embodiment of a conversion decoding circuit that performs color conversion and performs decoding. The difference from the circuit for performing the brightness conversion in FIG. 8 is that a color data conversion unit 100 is provided instead of the brightness conversion unit 322. The amount instructing how to convert the color data is input to the color data conversion unit 100 via the color conversion amount control line 101. The converted color information is C '
Becomes FIG. 14 shows a circuit example in the case where the color data conversion unit 100 is configured by a look-up table. Next, a color conversion method is performed as shown in FIGS. 15 (a) and 15 (b). The conversion amount data is divided into 8 bits. Of the 8 bits, the hue θ is determined by the above 4 bits, and the saturation h is determined by the lower 4 bits.
Gives the amount of change. <Modification> In the embodiment described above, L ^* a ^* b ^* is used, but L ^* u ^* v ^* , or NTSC YIQ, PAL, or YUV can also be used. The orthogonal transform is shown by Hadamard transform, but discrete CO
S conversion or slant conversion is also possible. Although the quantizer is described as vector quantization, it is not particularly limited. Note that the bit distribution of the brightness information L, the structure information S, and the color information C is not limited to that described in the above embodiment. Further, the input signal is not limited to RGB, and input of YGC or the like may be considered depending on the sensor. Further, in FIG. 1, a ^* b ^* is represented by an average value, but it may be stored in detail when it is available. In the examples, C
The expression of (color information) is represented by hue + saturation, but is not limited to this. [Effects of the Invention] As described above, according to the present invention, since color image data is divided into information on brightness, structure, and color and encoded, efficient encoding can be performed. On the other hand, the quantization rate at the time of quantizing the plurality of pieces of information (brightness information, structure information, and color information) is set in the order of brightness information> color information> structure information. It is possible to perform suppressed and efficient encoding. That is, it is possible to achieve both efficiency of compression and maintenance of image quality.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a configuration diagram of an encoding circuit according to an embodiment of the present invention, FIG. 2 is a diagram illustrating a configuration of color image data encoded by encoding according to the embodiment, FIG. FIGS. 3 and 4 are diagrams each showing a part to be processed in brightness conversion and color conversion. FIG. 5 (a) is a diagram showing an example of color image data in the embodiment, and FIG. 5 (b). (C) is a diagram when the edge portion of the color image data is represented by RGB or L ^* a ^* b ^* , FIG. 6 is a configuration diagram of the L ^* a ^* b ^* conversion unit, and FIG. 8 and 9 are circuit diagrams for performing brightness conversion and decoding, FIG. 10 is a diagram showing a conversion example of brightness conversion, and FIG. 11 is a diagram showing a configuration of color information C. FIGS. 12 (a) and 12 (b) are diagrams showing the relationship between hue θ and saturation h in the color information, and FIGS. 13 and 14 are diagrams showing decoding by performing color conversion. 15 (a) and 15 (b) are diagrams showing conversion examples of color data conversion. In the figure, 100... Color data converter, 302... L ^* a ^* b ^* converter,
304: orthogonal transformation unit, 307, 309, 313: quantization unit, 311
... a ^* average part, 312 ... b ^* average part, 315 ... MUX.

Claims (1)

(57) [Claims] Division means for dividing the color image data into blocks; first extraction means for extracting lightness information related to lightness and color information related to colors from the blocks; and orthogonal transformation of the lightness information A second extracting means for extracting the average brightness information in the block and the structure information of the brightness in the block, and extracting the average color information in the block from the color information; A quantizing means for compressing the quantity,
The extracted average brightness information, structure information, and average color information are quantized by setting quantization factor for average brightness information> quantization ratio for average color information> quantization ratio for structure information of brightness. A color image data encoding device, comprising: a quantization unit.