JPH02202775A - Encoding and decoding method for color picture data - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention] [Industrial application field]

The present invention relates to a method of encoding and decoding color image data, and particularly to a method of encoding and decoding color image data suitable for use when compressing and encoding color image data. .

[Conventional technology]

Conventionally, when encoding and compressing image data for printing, a YIQ variation method, which is known as a band compression technique for images typified by television (TV), has been applied. Note that this YIQ conversion method converts the color information of the three primary colors red (R), green (G), and blue (B) into color information based on the chromaticity axes of luminance Y and chromaticity ■ and Q, which have a lower signal amount correlation. This is a method of converting. When applying this YIQ conversion method to encode image data for printing, first, we wanted to invert the image data of cyan (C), magenta (M), and yellow (Y), which were regarded as complementary colors) , G, and B, YIQ conversion is performed using the following conversion formula (1), which is equivalent to the above-mentioned television. ... (1) Next, considering the resolution characteristics of human vision <m degrees (Y requires high resolution, the resolution for chromaticities ■ and Q is relatively low), the chromaticity 11Q By thinning out the pixels of the image data, the encoding efficiency was improved.

[Problem to be solved by the invention]

By the way, in the conventional printing image data encoding technology, the target of chromaticity axis conversion by YIQ conversion is limited to printing image data of <C, MSY- as shown in equation (1) above. . However, the printing image data includes black (K) image data in addition to C, M, and E, and in the conventional technology, the K image data is not subject to the YIQ conversion. It was only possible to encode and compress image data by treating it as a monochrome image. That is, it cannot be said that YIQ conversion alone can convert the printing image data of four colors of C, M, and Y-1K into a form with lower signal amount correlation. Therefore, the conventional technique has a problem in that printing image data cannot be efficiently encoded and compressed. The present invention has been made to solve the above-mentioned conventional problems, and is a code for color image data that can convert printing image data into a form with lower signal amount correlation, and can efficiently encode and compress the image data. The first step is to provide a method for
This will be the subject of this study. A second object of the present invention is to provide a decoding method that can reliably restore color image data compressed by the encoding method to the original data.

[Means to achieve the task]

The present invention is applicable to cyan (C), magenta (M), yellow (Y).
), a method of converting and encoding black (K) print image data on a chromaticity axis, the cyan, magenta, and yellow image data being converted into image data on a chromaticity axis including luminance Y1, Predict black image data from the converted image data of luminance Y, calculate the error of the predicted image data with respect to the actual image data of black, and calculate the image data of the chromaticity axis including the converted luminance Y and the calculated image data. The first problem is achieved by encoding the error. Further, the present invention decodes the image data and error of the chromaticity axis including the luminance Y encoded as described above, and obtains the decoded luminance Y.
predict black image data from the image data of , to yellow image data, and restore the printing image data of cyan, magenta, yellow, and black, thereby achieving the second problem.

[Operation and effects of the invention]

In the present invention, when converting and encoding the chromaticity axes of C, M, Y-1K print image data, C, M, Y- image data is more closely correlated with signal amount among the image data. The chromaticity axis is converted into image data of a chromaticity axis including a low luminance Y, for example, a luminance Y1 chromaticity (2), Q. Next, K image data is predicted from the converted image data of luminance Y, and an error eK of the predicted image data with respect to the actual K image data is determined. This determined error eK has a much smaller variance than K. Therefore, by encoding the image data of the converted YIQ and prediction error eK, it is possible to efficiently compress the printing image data of the four colors C, M, Y, and K in a form with lower signal amount correlation. Can be done. In addition, since image data can be processed by line sequential processing, the amount of memory can be reduced compared to the blocking method. Furthermore, since the conversion can be performed using a relatively simple algorithm, it is easy to implement the conversion into hardware. Further, since the signal amount correlation is lowered and the compression efficiency is improved, excellent effects such as higher image quality can be obtained at the same compression rate. In addition,
For example, the DPCM method can be used to encode the compressed YIQeK signal. In this way, encoding can be performed with a relatively simple conversion algorithm. Further, in the present invention, when restoring the image data for printing from the image data of the chromaticity axis including the encoded luminance Y as described above and the error, the chromaticity axis including the encoded luminance Y decode the image data and the error, predict black image data from the decoded image data of luminance Y, restore the actual black image data from the predicted image data and the decoded error, and Chromaticity axis image data including luminance Y is converted into cyan, magenta, and yellow image data to restore the printing image data. Therefore, it is possible to reliably restore encoded image data when transmitting print image data.

【Example】

Embodiments of the present invention will be described in detail below with reference to the drawings. First, a first example will be described. This first embodiment is an apparatus that converts 01M and Y-1K printing image data into YIQeK data based on the procedure shown in FIG. 1, and encodes the converted YIQeK data. The overall configuration of this device is shown in FIG. As shown in FIG. 2, this device includes a scanner 2 for capturing C, M, and Y-1K image data from a color transparent original for printing 1, and a C, M, and Y-1K image data output from the scanner 2.
A first memory table 3 for converting Y- image data into R, G, B image data, and converted R, G,
A second memory table 4 for converting the image data of B into image data of luminance Y1 chromaticity I, Q, and the converted chromaticity 1. A thinning section 5 for thinning out pixels for the image data of Q, and a thinned out chromaticity 1. Q, and a linear quantization unit 6 for linearly quantizing the data of the prediction error eK described later, and a linear quantization unit 6 for linearly quantizing the data of the image data of the luminance Y and the image data of the linearly quantized chromaticity 11Q and the prediction error eK.
An encoding unit 7 for encoding using the M method; and an encoding unit 7
A prediction unit 8 that locally decodes image data of luminance Y encoded in the image data and predicts K using the decoded luminance Y value; value and
It mainly consists of an addition point 9 for calculating the error (prediction error eK) from the predicted value of K (K') and inputting it to the encoding section 7. Note that the encoded image data is output from the encoding unit 7 and transmitted, or is transmitted to a recording medium 30 (described later).
is accumulated in As shown in FIG. 3, conversion tables 10, 11, and 12 are written in the first memory table 3 for converting the image data of 01M and Y- into each image data of RlGlB. C, M. C, M,
Y- image data is converted into R, G, and B image data. The second memory table 4 includes a conversion table 14.15.1 for converting RlG, B image data into image data of luminance Y1 chromaticity IQ, as shown in FIG. 4(A).
6 is written, and the R, G, B image data is converted into YIQ image data. The encoding unit 7 performs encoding using the DPCM method, and as shown in FIG. 5, in order to quantize the difference value e1 described later, linear quantization is performed in the first stage and maximum quantization is performed in the second stage. Quantizer 22 and the difference value q(e+) after quantization
is locally decoded to obtain the pixel value x rJ , delayed by one pixel, and the pixel value already decoded and delayed in the previous pixel x t1 The difference from the actual pixel value x1 (difference value e+) A local decoder 24 is used to calculate the quantized difference value q(e+), and a one-dimensional previous value prediction is used to multiply the quantized difference value q(e+) by a co-measurement coefficient a. can be taken as 1. If the encoding unit 7 is configured under such conditions, the configuration of the arithmetic circuit and the like will be simplified and economical efficiency will be improved. The input/output state of signals before and after the prediction section 8 is as shown in FIG. That is, as shown in FIG. K's actual data and predictions/i! [The difference between K' is calculated and the difference is input to the linear quantization unit 6 as the prediction error eK. Incidentally, a recording medium 30 for storing encoded image data as shown in FIG. 1 is connected to the encoding section 7. As this recording medium, for example, a hard disk, magnetic tape (MT), streamer (small MT), or optical disk can be used. The operation of the first embodiment will be explained below. In this first embodiment, based on the procedure shown in FIG. 1, C and M images are read from a color transparent original 1 shown in FIG.
, Y-1K printing image data, YIQeK de, -
The converted data is then encoded and compressed. When this procedure starts, first, C, M, E, and K image data signals are inputted into the first memory table 3 from a color scanner or the like (Pl). Among the input image data, 01M, Y- image data is converted into R, G, B image data (P2). At this time, it is assumed that C, MlY and RlG, B are complementary colors to each other, and the conversion is performed using the following conversion formula (2). In the embodiment, the conversion result of equation (2) is written in advance.
The LtJT format for accessing the first memory table 3 shown in FIGS. 2 and 3 is used. Next, the image data converted into R, G, and B is input into the second memory table 4, and the luminance Y and chromaticity 1. Q's YI
Convert to Q data (P3) This allows R, G, and B data to be converted to YIQ data, which has a weaker signal amount correlation. In the embodiment, the conversion result of equation (3) is written in advance.
The processing time is reduced by using the LUT format for accessing the second memory table 4 shown in FIGS. 2 and 3. Next, in order to improve the encoding efficiency of the image data converted to YIQ, the thinning unit 5 extracts the chromaticity I,
Regarding the data of Q, pixels are thinned out (P4). In this way, the reason for thinning out the chromaticity I and Q data without thinning out the luminance Y data is that the resolution characteristics of human vision are
It is known that the chromaticity ISQ is high, but the chromaticity I and Q are relatively low. Therefore, even if the chromaticity ISQ is appropriately thinned out, there is little effect on image quality. Furthermore, the compression rate can be changed by adjusting the thinning rate. In the embodiment, as shown in FIG. 7, for image data of chromaticities l and Q, image data is thinned out every other pixel in the main scanning line direction and sub-scanning line direction, and the luminance Y is thinned out. Image data is not thinned out. Next, the linear quantization section 6 performs linear quantization on the thinned out image data of chromaticity 2 and Q (P5). Normally, this linear quantization is performed using 8 bits from 0 to 255, but if it is desired to further improve the compression efficiency, linear quantization may be performed using 8 bits or less. Next, image data of luminance Y converted from CMY- and linearly quantized chromaticity 1. The image data of Q is encoded by the DPCM method in the encoding unit 7 shown in detail in FIG. 5 (P6). That is, in this encoding section 7, first, the local decoder 2
4 has already predicted the pixel value Xi by the previous value prediction which multiplies the previous restored pixel value Xi-1' on the same scanning line by the prediction coefficient a, and then 1, \
, find the difference (difference value) between the predicted pixel value x 1 and the input pixel value x 1 and convert the pixel value Xi into a difference numerator jller r. The variance of this difference value e1 is 1/10 or less of the variance of pixel value x 1 of image data, making it possible to transmit and store a small amount of data. Next, the converted difference value e1 is quantized by the componentizer 22. In this case, the quantization of the difference value e1 is performed in two stages: a first stage and a second stage. In the first stage, linear quantization is performed to reduce 9-bit data to data of 9 bits or less, and then, in the second stage, max quantization is performed to optimally quantize the difference value eI to create quantized data. a
(e+) and minimize the average embellished noise power. This max quantization is a method in which the range with a statistically high probability of occurrence is quantized finely, and the range with a low probability of occurrence is quantized coarsely, and the quantized data q(e+) is created assuming that the error distribution is a Laplace distribution. Alternatively, quantized data q(e+) is created from the actual data of the error distribution. Next, the quantized difference value Q(ei) is sent to the encoder 26.
The coded image data is created by encoding the image data, and the image data based on the code is used for transmission or storage. In this encoding, the minimum average code length can be achieved by using a Huffman code created according to the prediction error distribution. Note that the pixel value x rl is restored in order to predict the next pixel value on the same scanning line from the quantized difference value q(e+). As described above, YIQ image data is encoded and compressed. Next, the encoding unit 7 locally decodes the luminance Y, and the decoded luminance Y is used to predict K in the predictor 8 shown in FIG. An error (prediction error) eK in the value of K is determined (P7). There is a strong correlation between image data of luminance Y and K. Therefore, various color images are subjected to YIQ conversion, the correlation between luminance Y and K is statistically investigated, and the prediction unit 8, which is converted into an LUT, predicts K based on the prediction data created from the investigation results. . To create LLIT in the prediction unit 8, as shown in FIG. 8, in order to determine the representative point (representative value) of K,
Determine the representative point (representative value) of luminance Y in advance, and
The distribution of K is measured for each representative point. Next, a representative point of K is determined from the measured distribution of K. When determining this representative point, the point with the maximum distribution frequency can be used as the representative point, or the average value, median value, etc. of the distribution may be used as the representative point. Next, from the obtained representative points of K, a prediction curve of K with respect to luminance Y is determined by -th order linear approximation, and a prediction formula of K is determined. From this prediction formula, K for the input value Y1 of luminance Y is
Write ' + to the predicted value of LU in the memory table.
Let it be T. Subtract the actual value of K from 'l to the predicted value of K obtained from this LtJT and predict 'i11! Find the error eK. Note that the variance of the calculated prediction error eK can be reduced to approximately ⅓ to ⅛ of the variance of the actual data of K. Thereafter, the prediction error eK is input to the linear quantization unit 6 and linearly quantized. In the example, this linear quantization is O~25
However, if it is desired to further improve the compression efficiency, the data can be quantized to 8 bits or less. The prediction error eK linearly quantized as above is expressed as YiQ
Similarly to the signal, the encoder 7 encodes the signal using the DPCM method (P8). Next, the encoded YIQeK image data is written into a predetermined i* typeface 30 and stored or transmitted (P9). Next, it is determined whether the processing of all pixels on the screen to be processed has been completed, and if not, the process returns to the beginning (Pl) of the procedure in Figure 1 and processes other pixels. (Plo) Processing is continued in this manner until all pixels on the screen are encoded. In addition, in the first embodiment, from CMY- to RGB
Although LLIT is used for the conversion, it is also possible to use a circuit that executes equation (2). Further, although an LUT is used for conversion from RGB to YIQ, it is also possible to use a circuit that calculates equation (3). Moreover, the conversion table 14 in the second memory table 4 is
It is also possible to implement the configuration as shown in FIG. 4(B). The conversion table 14 in FIG. 4(B) is an LUT of the floating point product operation (for example, anXR, etc.) regarding Y when the matrix operation of YIQ conversion is expanded as shown in the following equation (4). . In this way, by using a configuration in which floating-point product operations, which generally require a long calculation time, are converted into LUTs, high-speed processing can be achieved and the memory capacity required for LLIT can be reduced. Note that the conversion tables 15 and 16 regarding I and Q in the second memory table 4 also perform the product operation of I and Q in equation (4) as shown in FIGS. 4(C) and (D), similarly to Y. It may be configured as an LUT. Next, a second example will be described. This second embodiment is an apparatus that decodes data encoded by the steps P1 to P10 (hereinafter referred to as encoded data) and performs various processes to restore it to image data for printing. As shown in FIG. 9, this device decodes input YIQeK encoded data and also decodes DPCM
a decoding/restoring unit 32 for restoring YIQeK image data using the method; a second prediction unit 34 for predicting K image data from the restored luminance Y image data;
A second addition point 36 for restoring K image data by subtracting the restored prediction error image data from the predicted K image data, and the restored chromaticity 1. The LUT includes an interpolation unit 38 for interpolating Q image data, and conversion tables 40A, 40B, and 140G for converting the restored and interpolated YIQ image data into R, G, and B image data.
a third memory table 40 for converting in the format, RlG,
and a fourth memory table 42 having a conversion table 42A1428, 420 for converting 8 image data into C, M, Y- image data and converting it in LUT format, and decoding and restoring the encoded data. , is executed according to the procedure shown in the flowchart of FIG. Note that the detailed configuration of the decoding/restoring unit 32 and the like of this device will be explained in conjunction with the explanation of each procedure below. That is, when the procedure shown in FIG. 10 starts, first, the encoded data is input to the decoding/restoring section 32 (DPI). The encoded data to be input is stored in a storage medium (hard disk, M
The data may be decoded data read from a T, streamer, optical disc, etc.) or transmitted decoded data. Next, among the input encoded data, Y, ■, and Q encoded data are decoded and restored using the DPCM method (DP2
). Next, the encoded data of the input prediction error eK is decoded and restored using the same DPCM method as in step DP2 (DP3). That is, in the decoding of the YIQeK encoded data in steps DP2 and DP3, first, the decoder 44 shown in FIG. This is done by decoding it as the difference value q(e+). In this case, the decoder 44 decodes using a code table corresponding to the Huffman code table used by the encoder 26 of FIG. 5. Next, the pixel value x+I of the YIQ image data is restored from the difference value q(e+) decoded in this manner. In this case, the pixel value xl+ is restored in advance of the previously restored pixel value Xt1' on the same scanning line. Here, the same value as the prediction coefficient value used in encoding by the encoder 26 is used as the prediction coefficient a. In the example, it is a −1. Next, in order to restore the image data of K, a predicted value (K') of K is first determined from the luminance Y data restored in step DP2 in the second prediction unit 34 shown in detail in FIG. 12. This second prediction unit 34 uses the same configuration and operation as the predictor 8 shown in FIG. 6 above, and the explanation thereof will be omitted. Next, by subtracting the prediction error eK decoded in step DP3 from ' to the prediction 1 of K at the second addition point 36, the image data of K is restored (DP4). Next, since the image data of chromaticity ■ and Q restored in step DP2 is in a thinned state, the thinned out image data is interpolated in the interpolation section 38 (DP5
). In this embodiment, the nearest neighbor interpolation method is used for this interpolation, and FIGS. 12(A) and 12(B) show examples of image data before and after interpolation using the nearest neighbor interpolation method. That is, the thinned image data before interpolation is shown in FIG. 12 (A
), the image data will be in the state shown in FIG. 12(B) after interpolation. Note that in the example, interpolation is performed using the nearest neighbor interpolation method, so
Since the processing is simple and high-speed processing is possible, it is easy to implement hardware. Also, the chromaticity 1. As a method for interpolating the image data of Q, in addition to the nearest neighbor interpolation method, a pi-linear interpolation method or a cubic interpolation method can be used. Next, the restored and interpolated YIQ image data is converted into RG
Convert to B image data (DP6). This conversion is performed using the conversion formula shown in the following equation (5). Note that the coefficient matrix of equation (5) is the inverse matrix of the coefficient matrix of equation (3) used in step P3 in FIG. (5) Normally, when conversion is performed by calculation, the processing speed is faster if an LIJT in which the conversion results are written in advance is used. Therefore, in this embodiment, a third memory table 40 is used, as shown in detail in FIG. The YIQ data is converted to RGB data using the LUT format that accesses the conversion tables 40A, 140B, and 400 of the third memory table 40. This improves efficiency by reducing processing time. Next, convert the converted RGB data into CMY using the following equation (6).
Convert to data (DP7). C-255-R C-255-R・・・・・・・・・(6) Y-255-
In the case of the B embodiment, a fourth memory table 42 as shown in detail in FIG. 15 is used, in which the conversion result of equation (6) for predetermined values is written in advance, and the RGB data is converted to CMy-data. When converting, the conversion table 42A142B of the fourth memory table is converted based on the RGB data.
, 42G is accessed in LUT format. This reduces processing time and improves efficiency. Through the processing of steps DP1 to DP8 above, CMYK image data is restored from the YIQeK encoded data, and this CMYK data is transferred to a secondary medium such as a hard disk or to an output device (scanner, cathode ray tube (CRT), color printer, etc.) (DP8). Next, it is determined whether the processing of the above steps DP1 to DP8 has been completed for the image data of all pixels (DP9),
If it is determined that the processing for all pixels has not yet been completed, the process returns to the first step DP1 again, and the procedures from this step DP1 onwards are executed for the data of other pixels. On the other hand, if it is determined that the processing of all pixels has been completed, the restoration of the encoded data is completed. In this way,
All pixels of the entire screen are restored to obtain image data for printing. Note that in the second embodiment, an LUT was used for conversion from YIQ to RGB, but it is also possible to use a circuit that calculates equation (5). Also, convert RGB data to YI
LIJT was used to convert to Q data, but (6
) It is also possible to use a circuit that calculates the equation. Moreover, the conversion table 40A in the third memory table is
It is also possible to implement the configuration as shown in FIG. 14(B). 1st
The conversion table 40A in FIG. 4(B) converts floating-point product operations (for example, finXY, etc.) regarding R when the inverse matrix operation of YIQ conversion is expanded as shown in the following equation (7) into LLI.
It is a T version. R= bnXY+ btzXI+ b13XQG =
bzt X Y + L□Xl+bt3XQ...
(7) B= b3tXY+ b32X I+ b33X
QBy using a configuration that converts floating point product operations, which generally require a long calculation time, into LUTs, high-speed processing can be achieved and the memory capacity required for LIJT can be reduced. Note that 40B regarding G and 8 in the third memory table
, 40C, like R, can be configured by converting the product operation of GlB in equation (7) into an LUT as shown in FIGS. 14(C) and 14(D). In the first actual dust example, the image data is changed from CMY to R.
Then, the chromaticity axis was converted to NTSC YIQ image data (P2). Further, in the second embodiment, the image data was converted from YIQ to RGB (DP6) and CMY (DP7). However, the chromaticity axis conversion when implementing the present invention is not limited to this. For example, instead of the YIQ chromaticity axis, L"a"b" lx u' vX, YLIV
(7) It may be converted to the chromaticity axis. Also, instead of the YIQ chromaticity axis data of the NTSC system, if the chromaticity axis includes luminance Y, other chromaticity axes, such as Y1 of the PAL system,
It can also be converted to Q data. Furthermore, in the first and second embodiments, the YIQeK data was encoded and compressed using the DPCM method, and also decoded, but the encoding and decoding technology is not limited to this DPCM method, and other methods may be used. Encoding and decoding can be performed using orthogonal transform, such as discrete cosine transform, Hadamard transform, and slant transform.

[Brief explanation of the drawing]

FIG. 1 is a flowchart showing a procedure for compressing and encoding image data according to a first embodiment of the present invention, FIG. 2 is a block diagram showing the overall configuration of the device of the embodiment, and FIG. FIGS. 4(A) to 4(D) are block diagrams showing details of the first memory table for converting image data from CMY to RGB used in the embodiment. FIG. 5 is a block diagram showing the configuration of the encoding section in detail. FIG. 6 is a block diagram showing the configuration of the K2OK converter. FIG. 7 is a block diagram showing the configuration of the K2OK converter. , Similarly, a plan view showing an example of thinning out image data of chromaticity ■ and Q, FIG. FIG. 9 is a block diagram showing the configuration of an apparatus for restoring encoded image data according to a second embodiment of the present invention, and FIG. A flowchart showing a procedure for restoring image data; FIG. 11 is a block diagram showing a schematic configuration of a restoration circuit for restoring encoded data and obtaining pixel data according to the second embodiment; FIG. FIG. 13 is a block diagram showing the basic configuration of a circuit for restoring image data of K.
14A to 14D are diagrams showing examples of pixels before and after interpolation of thinned image data for explaining the operation of the embodiment. FIG. 15 is a block diagram showing the structure of the fourth memory table. FIG. 15 is a block diagram showing the structure of the fourth memory table. 2... Scanner, 3... First memory table, 4... Second memory table, 5... Thinning section, 6... Linear quantization section, 7... Encoding section, 8... Prediction unit, 9... Addition point, 36... Second addition point, 38... Interpolation unit, 40... Third memory table, 40A, 40B, 40C... YiQ →RGB conversion table, 42... Fourth memory table, 42A142B, 42C... RGB-CMY conversion table, 44... Decoder. 10.11, 12...CMY-→RGB conversion table, 14, 15,
16... RGB-4YIQ conversion table, 22... quantizer, 24... local decoder, 26... coder, 30... recording medium, 32... decoding/restoration part, 34... second prediction part,

Claims (2)

[Claims] (1) A method of encoding cyan, magenta, yellow, and black printing image data by converting the chromaticity axis, and converting the cyan, magenta, and yellow image data into chromaticity axis image data including luminance Y. predict black image data from the converted image data of luminance Y, calculate the error of the predicted image data with respect to the actual image data of black, and calculate the image data of the chromaticity axis including the converted luminance Y. and a method for encoding color image data, characterized in that the obtained error is encoded. (2) Decoding the image data and error of the chromaticity axis including luminance Y encoded by the encoding method of claim 1, predicting black image data from the decoded image data of luminance Y, and predicting. Restore the actual black image data from the image data and the decoded error, convert the decoded chromaticity axis image data including the luminance Y into cyan, magenta, and yellow image data, and then convert the image data to cyan, magenta, and yellow. , yellow, and black printing image data.