JPS59138183A - System for coding color picture - Google Patents

Abstract

PURPOSE:To reduce the amount of data of a code representing run color and the boundary position of the run by referencing the group of the run before a run to be converted and that in a scanning line above the scanning line including said run to be converted. CONSTITUTION:A run detecting circuit 3 detects the run from a picture data inputted in the order of scanning lines, outputs the changing position of run (absolute) and the run color to a memory 4, a position coding method designating circuit 5 and a color forecasting circuit 6 respectively and outputs the run color to a coding circuit 7. The position coding method designating circuit 5 receives an input of the changing position to be coded, references the changing point position and the run color stored in the memory 4, selects the coding method of the changing point position by a prescribed method and outputs the selected position coding method to the coding circuit 7. The color forecasting circuit 6 receives the input of said run color and the changing point position (*) to be coded, references the changing point position and the run color stored in the memory 4, forecasts the run color in the prescribed color forecasting method, compares the forecast value of the color with the actual value, and outputs a signal representing whether or not they are identical to the circut 7.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a coding method for efficiently transmitting and storing color images.

Colored images, in which two-dimensional patterns composed of characters and figures are colored with multiple colors, have a significantly greater effect as visual information than black and white images, and are therefore used in various applications such as teletext broadcasting and conversational text and graphic information systems. It is expected that its use in image information systems will expand. However, when attempting to transmit or store such three-color images, the required transmission time and storage capacity become larger than those for monochrome images. To deal with this problem, various methods have been considered for compressing data by encoding color images, and a typical example is a color run-length encoding method.

However, even with this method, the data compression effect is insufficient if the color image is complexly colored.
Furthermore, a method may be adopted in which restrictions are placed on the coloring itself of the color image. For example, a color image is divided into a large number of small rectangular areas, and each small rectangular area is restricted from using two or more colors. Although such restrictions are convenient for data compression, they impede the ability to express images. Because the restrictions are artificial,
Since normally created illustrations and graphs cannot be transmitted as they are, creating color images as information requires extra effort and ingenuity, and completely avoids unnatural coloring. New problems will arise, such as the inability to do so.

It is an object of the present invention to provide a color image encoding method that compresses color images more efficiently than conventional methods and eliminates the need to limit the image representation ability as described above.

Hereinafter, the color image encoding method according to the present invention will be explained in detail with reference to the drawings.

In the following explanation, for the sake of simplicity, we will explain how to transmit (i.e. spatially move) and store (i.e. spatially move) a color image.
(that is, moving in time) are collectively referred to simply as transmission. Furthermore, the encoding of a color image refers to the process of converting an image into color information for each pixel (hereinafter referred to as image data), which is decomposed into color information for each pixel by normal rask scanning, into code data equivalent to the color information. What is image decoding?
It means the process of inverse conversion from encoded data to image data. Regarding color information, we consider two elements: hue (the types of the eight primary colors and their mixtures) and gradation (the degree of lightness and darkness for each hue), and the combination of these two elements is simply called a "color."I'll decide.

Furthermore, in the following explanation, it is assumed that image data is converted into code data in scanning order, and the pixels included in each run are focused on a series of pixels having the same color within a scanning line (hereinafter referred to as a run). Let us consider an encoding method in which the color of each run (hereinafter referred to as run color) and the position within the scanning line of each run are represented by a code. The position of the run within the scan line is the left or rightmost pixel of the run (i.e. the pixel at the border of the run)
is expressed by where is located within the scan line,
Rather than expressing the position of a run boundary pixel within a scan line using an absolute position such as the number of pixels from the left end of the scan line, it is better to express the position of a run boundary pixel in a scan line that is already known (that is, already encoded). How many pixels in the scan line direction from the boundary pixel of the run (
However, it is known that the statistical distribution is more concentrated and the encoding efficiency is better if the relative position is expressed as deviated (with rightward direction being positive and leftward direction being negative). In the following, the leftmost pixel of the run will be called a change point, and the position of each change point will be expressed as a relative position from other change points whose positions are already known, and this will be encoded as the position of the run. do.

Considering the above points, we will first review the conventional color run length encoding method.

Figure 1 is a diagram showing an example of encoding using the color run-length encoding method, and Figure (a) shows one piece of image data to be encoded.
A schematic diagram of a scanning line segment and a code format obtained by encoding the scanning line segment are shown in FIG. In Figure (a), when there is a change point as shown by the arrow in the image data (assuming that there is also a change point on the right side of the scanning line),
The run color of the i-th run from the left in the scanning line is expressed as C(i), and the relative position of the change point of the (i+1)-th run to the change point of the i-th run is expressed as R(i). There is. Here, the run colors of adjacent runs are different, and C(i)≠C
(i+1).

Furthermore, all relative positions R(i) are positive integers, which are equal to the number of face trees included in the i-th run (this is called a run length). Then, by encoding, each run is converted into a code representing the run color and a code representing the run length, as shown in Figure (b).

Note that "EOL" in FIG. 3B is a code indicating a separation between scanning lines. The code representing the run color may be a fixed length code that directly represents the color element by element, as shown in the example shown in Figure 1 (C), or a non-fixed length code determined based on the probability of appearance of the run color. It can also be a sign. In addition, codes representing run lengths include modified Huffman codes (hereinafter abbreviated as MH codes), which are well known as binary image data encoding methods, and non-constant length codes such as Wyle codes. Usually, a symbol is used. Here, it is obvious that in binary image data such as a black and white facsimile signal, two colors appear alternately for each run, so the code representing the run color can be omitted, but in the color image according to this invention, each color Note that a code representing the run color of the run is required. It goes without saying that the code representing the run color and the code representing the run length must be distinguishable from each other based on the transmission order of the codes and the bit arrangement within each code word.

As can be seen from the above explanation regarding Figure 1, in the conventional color run-length encoding method, encoding is performed by focusing on a run within a scanning line, but if the color changes even slightly in the scanning order, it must be treated as a different run. Moreover, since it is necessary to add a code representing the run color to each run, the amount of codes increases compared to a black and white binary image. Also, although the relative position of the change point is encoded, it is a simple method that uses only the relative position to the immediately previous change point within the scan line, so it is not possible to encode the correlation between scan lines that the image originally has. is not utilized at all.

The color image encoding method according to the present invention is intended to improve the above-mentioned drawbacks of the color run-length encoding method and realize more efficient data compression. Therefore, in the color image encoding method according to the present invention, within the scanning line including the run that is about to be converted into a code (the relevant run), the NU of the run to the left of the run and the NU of the run above the scanning line including the relevant run. The color of the run is predicted by the method described later by referring to the group V of runs in the scanning line, and if the prediction is correct, the code representing the run color of the run is omitted, and the position of the run is also described later. The color image is converted into a code representing the relative position (in the scanning line direction) of the boundary pixel of the run with respect to the boundary pixel of the run that belongs to run group V or run group U and satisfies a predetermined condition. The two-dimensional correlation between run color and run position is actively utilized.

FIG. 2 is a diagram showing the encoding principle in the color image encoding method according to the present invention, and FIG. 2(a) shows one scanning line segment of image data to be encoded and one scanning line segment immediately above it. What is schematically illustrated and FIG. 2(b) shows the format of the code obtained by encoding the image data of FIG. 2(a). The arrow in figure (a) indicates the change point, and the run color of the i-th run from the left in the scanning line to be encoded is C(i), and the (i+1)-th run color for the known change point position is C(i). Let A(i) be the relative position of the change point of the run. and C(i)
The symbols representing A(i) and A(i) are arranged as shown in Figure (b) and
Image data for scanning lines will be expressed. At this time, the code of the run color C(i) and the code representing the relative position A(i) must be distinguishable from each other. Here A
In principle, (i) represents the relative position to the change point closest to the change point of the run among the change points whose positions are known within the two scan lines. At this time, an identification code is added as necessary so that the relative position from which point of change can be identified. In the example of Figure 2(b), A(1), A(3)
, A(N-1) are all relative positions from the change point in the scanning line directly above, and only A(2) is relative position from the change point in the same scanning line. Therefore, A(
An identification code (denoted as HRz) is added before 2).

Regarding run color C(i), we assume that all run colors except run color C(2) of the second run can be predicted correctly.
1).

The symbols representing C(3), . . . C(N-1) are omitted.

As can be seen from the above principle, the more similar the positions of the change points of the run are in the upper and lower scanning lines and the more concentrated the distribution of relative positions, the greater the probability that the prediction of the run color is correct.
The code length of the encoding method according to the present invention is shortened, and the degree of improvement in encoding efficiency is greater than that of the conventional color run length encoding method.

FIG. 8 is a block diagram showing an embodiment of the configuration of the encoding circuit and decoding circuit of the color image encoding method according to the present invention, in which (1) is the color image encoding circuit, and (2) is the color image decoding circuit. (3) is a run detection circuit that detects the run and the run color; (4) is the memory that stores the change point position (absolute) of the run and the run color; (5) is the position of the change point of the run. (6) is a color prediction circuit that predicts the color of the run and compares the predicted value with the actual value; (7) is the position of the change point of the run (relative) (8) is a code identification circuit that identifies the type of code word to be input; (9) is a change point code. αQ is a position code decoding circuit that converts a position code into a change point position (relative) and further converts it into an absolute position by referring to a known change point position stored in the memory @. Color prediction circuit that predicts color, 0◇ is color code decoding circuit that converts run color code to run color, @ is decoded change point position (absolute)
and a memory for storing the run color (13 is an image data restoration circuit that converts the decoded change point position and run color into image data that can be displayed or recorded on the output device.The color image code shown in FIG. 8) In the conversion circuit (1), a run detection circuit (3) detects a run from image data input in the order of scanning lines, and stores the change point position (absolute) and run color of the run (4) in one-position encoding method. The output is sent to the designation circuit (5) and the color prediction circuit (6), and the run color is sent to the encoding circuit (
7). The memory (4) is a position specification method selection circuit (
5) and stores the change point position and run color to be referred to in the color prediction circuit (6). The change point position and run color referenced in the position encoding method designation circuit (5) and color prediction circuit (6) are included in the scanning line that includes the run to be encoded, and are included in the scan line that includes the run that is to be encoded. This is the changing point position and run color of the run included in the upper scanning line of the containing scanning line, and when the encoding of one scanning line is completed, the scanning line that was the scanning line containing the run until then becomes the upper scanning line number. ζ, the memory (4) is configured to update the changing point position and run color of the scanning line to be referenced for each scanning line. The position encoding method designation circuit (5) receives the input of the changing point position to be encoded, refers to the changing point position and run color stored in the memory (4), and determines the encoding method of the changing point position by the method described later. is selected, and the selected position encoding method is output to the encoding circuit (7). The color prediction circuit (6) receives input of the color and change point position of the encoded run, and performs color prediction described later by referring to the change point position t, Y and run color stored in the memory (4). The method predicts a run color, compares the predicted value of the color with the actual value, and outputs a signal to the encoding circuit (7) indicating whether the predicted value and the actual value of the color are the same. The encoding circuit (7) receives input of the position encoding method, the color of the run, and the comparison result between the predicted value and the actual value of the color, and uses the specified position encoding method for the change point positions j, i, and l. For run colors, when the predicted color value and the actual value are the same, no code word is generated, and when the predicted color value and the actual value are different, a run color code is generated and the change point position is determined. appended to the codeword. Color image decoding circuit (2
), the code identification circuit (8) identifies input in which a code word at a change point position and a code word for a run color are mixed, outputs a code word at a change point position to a position code decoding circuit (9), and The code word of the run color is output to the color code decoding circuit θ1), and for a run to which no color code is added, a signal indicating that the run color should be determined by prediction is output to the color prediction circuit α0. The position code decoding circuit (9) inversely converts the change point position code to obtain the relative position, and further refers to the already decoded (absolute) change point position in the memory a and inputs the change point position code. Calculate the change point position of the run represented by the position code. The color code decoding circuit (11) reversely converts the input color code word into color information. When the color prediction circuit 01 is instructed by the code identification circuit (8) to obtain the run color by prediction, the color prediction circuit 01 receives the decoded change point position from the position code decoding circuit (9), and also inputs the decoded change point position from the position code decoding circuit (9). Referring to the change point position and run color that have already been decoded in the memory (2), the color of the run is reproduced using the same method as the color prediction circuit (6) of the color image encoding circuit (1). Output. Memory @ stores the (absolute) change point position decoded by the position code decoding circuit (9) and the color decoded by the color code decoding circuit (1]) or the color restored by the color prediction circuit 0Q. The contents are referenced by the position code decoding circuit (9) and the color prediction circuit.

This memory CH5 is a memory having the same function and configuration as the memory (4) of the color image encoding circuit (1).

The image data restoration circuit α frame is a circuit that restores image data that can be displayed or recorded on an output device by reproducing color information for each pixel based on the decoded displacement point position and run color.

Next, a characteristic portion of the processing for color image encoding and decoding executed by the circuit configured as described above will be described in detail.

First, a method of encoding the change point position will be explained. FIG. 4 is a diagram showing the relative positional relationship between the run to be encoded and its neighboring runs.
It is written in the figure as .

The changing points for specifying the position of run P are the changing point of run P itself (this is referred to as a) and the changing point of the run to the right of run P (this is referred to as al). Since ao should be known by encoding the position of the run to the left of run P, the position of change point a1 should be specified by the code.

Also, at the change point in the scan line directly above the scan line that includes run P,
The position in the scanning direction is a change point a. The changing points on the right side are indicated in the figure as b1* b2 * b3... in order from the left.

When encoding the position of the changing point a1, use the relative position in the scanning line direction from any of the changing points bl * b2 + b3... in the scanning line immediately above (this is called vertical mode). If there is no appropriate change point to refer to in the scanning line immediately above, use the relative position from the change point aQ (that is, the normal run length) (this is called horizontal mode). Make it. In particular, if the change point in the scanning line directly above is within a predetermined range in the scanning line direction from the change point a1, the change point a is always at a relative position from one of them.
Specify the position of l. In Figure 4, the above range is a
It is illustrated within 10 pixels from left to right.

FIG. 5 is a diagram for explaining an example of a change point position encoding method by classifying the above-mentioned relative positional relationships.

FIG. 5(a) shows the change point b within a predetermined range from the change point a1.
1 exists, and in this case, the position of al is specified by the relative position to bl. For example, when al is shifted to the left by one pixel in the scanning line direction with respect to , it is specified by the code "one pixel to the left in vertical mode". In the figure, this is expressed as VL(1).

FIG. 5(b) shows a case where there is a change point bi (i≠1) within a predetermined range of change points al, and in this case, among change points b+ within the predetermined range, 31 closest to (a
a depending on the relative position to the change point (including the case where it is at the same position as l)
Specify the position of l. For example, in the example shown in Figure (b), there are b2 and b3 within the range, and only b2 is on the left side of al, indicating "the same position as the second change point in vertical mode". Therefore, in order to clearly indicate that the i-th change point is to be referred to, a code (indicated in the figure) indicating passing each of the (i-1) change points before it (this is called a pass mode) is used. P.A.
After (i-1) pieces of SS (denoted as S S) are connected, a code (denoted as VR(2)) of "two pixels on the right in vertical mode" is generated. In FIG. 5(c), there is no change point in the scanning line immediately above within a predetermined range from the change point a1, but there are i change points b1.
:... This is a case where bI exists. The position of h al is specified by the relative position from bl of the closest change point to the left of al. For this reason, first, after (j-1) PASS codes to clearly indicate that the i-th change point is referred to, a code indicating that the mode is vertical mode (denoted as VRT) is used.
is added, and a code representing the distance in the scanning line direction from bl to 81 (this is expressed as M(al-bi)) is added to the combination 6. Here, M(al-bI) has M
A code conventionally used as a run-length code such as an H code is used. Furthermore, if at least one PASS code precedes it (i.e., i≧2), it is obvious that the mode is vertical mode, so the code VRT may be omitted. FIG. 5(d) shows that there is no change point in the scanning line immediately above within a predetermined range from the change point a, and also from the change point a0 to a
This shows a case where there is no change point in the scanning line directly above even between 1 and 1. In this case, the position of al is a.

Specify by relative position from. After the code indicating that this is horizontal mode (this is written as HRz),
A code M (a
It is expressed by adding taQ ).

FIG. 6 is a schematic flowchart showing the procedure for encoding one change point in the above-described change point position encoding method.

The meaning of this flowchart corresponds directly to the explanation of FIG. 5 above. This flowchart should be executed mainly by the position encoding method designation circuit (5) in FIG. When it is necessary to generate the change point position code,
A code generation command is issued from the position encoding method designating circuit (5) to the encoding circuit (7).

Further, FIG. 7 shows a specific example of the types of codes required when the number is 8 in the above description. In this example, the change point position code and the run color code are non-fixed length codes that can be distinguished from each other. (Cp in the run color code is a constant length code as shown in Figure 1 (3)) The example of the change point position encoding method explained in Figures 5 to 6 and Figure 7 above is based on the scanning Each time a change point appears sequentially in the line direction, the mode of the change point is determined and a corresponding symbol is generated.Each change point b+ is processed individually, and finally the change point al is processing to finish encoding the position of one run.

Classification of the relative positional relationship of change points during change point position encoding is not limited to the above example; This is a diagram for explaining an example. In this example, the relative positional relationship is 8
It is classified as

Symbol a in Figure 8. 4. Reference characters b2 to b2 indicate the change points of the run explained in FIG. As shown in Figure 8(a), the change point a
1 is to the right of the change point b2, a pass mode code word (PASS) is generated as a pass mode, and the pixel ao' immediately below the change point b2 (not necessarily the change point)
A new a. , and perform the following encoding. (In other words, bl*b2*al is found based on the virtual a, and the next encoding is performed.) However, pixel a. When ' and the change point al are the same pixel, the pass mode is not established. As shown in FIG. 8(b), when the change point b1 exists within a predetermined range from the change point a1 (for example, within 'ro pixels from the change point a1 in the left and right scanning line direction), the change point al is changed as a vertical mode. It is encoded by the relative position from point b1. When To is 8, 7 codewords are required in vertical mode.

As shown in FIG. 8(C), within a predetermined range (
Same range as in vertical mode. ) at the change point b
1 does not exist, the horizontal mode is set and the change point a1 is changed to the change point a. Encoded by relative position from .

The code word for the horizontal mode is a code word (HRz) indicating the horizontal mode, and the relative distance between the change point a (, and the change point 81) are encoded using an MH code or the like.

FIG. 9 is a flowchart showing the procedure for encoding one change point in the change point position encoding method as shown in FIG. This flowchart is also mainly executed by the position encoding method designation circuit (5). Further, FIG. 10 shows a specific example in which the TO of code word types required in the example of the above-mentioned encoding method is set to 8. These code words are distinguishable from each other.

In the changing point position encoding method explained with reference to FIGS. 8 to 9 and 10, by imagining the changing point in the pass mode, the changing point in the scanning line directly above can be calculated by considering only bl and b2. good.

As can be seen from the explanation of the above two examples, there may be various other variations in encoding the change point positions depending on the method of classifying the relative positional relationship of the change points and the specific encoding procedure. For example, in vertical mode, instead of encoding the relative position of the leftmost change point b+ and change point al within a predetermined range, the relative position of change point bl closest to change point al within a predetermined range is encoded. You can also do this. Further, the value To that provides the predetermined range should be appropriately selected depending on the statistical properties of the image to be encoded, and furthermore, the predetermined range does not necessarily have to be symmetrical. Further, the change point in the upper scanning line does not have to be limited to one scanning line directly above.

Next, a method for predicting the color of an orchid will be explained.

In the following description, the color of run X will be expressed as Cx, the gradation of run X will be expressed as Yz+the hue of run X will be expressed as hue HX, and their predicted values will be expressed as predicted value C'X+predicted value Y'x, respectively.

It is abbreviated as predicted value H%. Further, the run for which the color or color element (gradation 2 hue) is to be predicted is run P1, the run to the left of the run P is run L, and the run to the right of the run P is run R. In addition, the relative distance between two pixels (pixel a and pixel b) is expressed as the number of pixels as before, and the value is added with a sign indicating that the right side in the scanning line direction is positive. Runs within a predetermined range defined by the distance from the run boundary position of run P within the scan line directly above the scan line are sequentially run from the left as run 1, run 2, and so on.

It is written like run M.

As an example of a color prediction method, an example of a color prediction method that combines gradation and hue will be described. FIG. 11 is a diagram for explaining an example of a color prediction method, and shows the relative positional relationship between the run P whose color is to be predicted and the runs in its vicinity that are referred to for prediction. . In this example, the run group V is limited to the run group within one scanning line immediately above the scanning line including the run P. Here, run group ■l is run l
The distance from the change point of the run P in a subset of v is −T,
A run group is defined as a run group including pixels whose distance from the change point of the run R on the right of the run P is within 10T2. That is, when predicting the color here, the run group v0 and the run group U are referred to. FIG. 11(a) shows a case where run 1 is the only run belonging to run group v1. In this case, the predicted color value C'P of the run P is assumed to be equal to color C1 when CL≠C1, and CL− In the case of C1, it is assumed that the color CK is equal to the color CK of the run adjacent to the left of the run L. FIG. 11(b) shows a case where there are two runs belonging to run group V1, run 1 and run 2,
At this time, the predicted value C'F- is assumed to be equal to the color q when CL=Q, and to the color c1 when cL = C2, and furthermore, CL-y! -c1 and CL9! =C2('
) sometimes run 1. Due to the relative positional relationship between run 2 and the run P, if the change point of run 2 is to the right of the center pixel of the run P, it is equal to color C, and the change point of run 2 is to the right of the center pixel of the run P. Color C when it is to the left of the pixel
Suppose it is equal to 2. FIG. 11(C) shows a case where there are 8 or more runs belonging to run group ■1, such as runs 1 to 4, and at this time, when one or more runs with color cL belong to run group V1. Then, the predicted value c is set to the color of the run for which the absolute value of the distance between the center pixel of the run and the center pixel of the run P is the smallest among the runs on the right side of the run with this color CL.
6 are equal, and when there is no run with color C in the run group ■, the predicted value G is the color Q of the run to the left of run L.
Suppose that it is equal to . FIG. 12 is a diagram showing an example of a schematic flowchart when the prediction explained in FIG. 11 is executed in the prediction circuit (6) and αO of FIG. 8 by a storage logic circuit including a microprocessor. ) and figure (b) are made to be continuous with each other. As can be seen from FIG. 12, the positional relationship between the run P whose color is to be predicted and the neighboring runs referred to for prediction is determined by a microprocessor or the like by determining the position within the scanning line of each run ( This can be determined by comparing the change point address). Further, calculation of the predicted value of the color of the run or its elements, comparison of the predicted value with the actual value, etc. can be easily executed by a microprocessor or the like.

This situation also applies to prediction methods other than the color prediction method explained in FIG. 11.

The prediction method described above is an example of a method of predicting the gradation and hue, which are color elements, all at once, but it is also possible to separate the color elements and predict each element.

An example of such a prediction method will be described next. In this example, first the gradation is predicted and then the hue is predicted. In this case, for runs encoded in vertical mode, gradation is predicted with a known run length and unknown hue, and hue is predicted with known run length and gradation, and encoded in horizontal mode. For runs to be converted, the gradation is predicted under conditions where the run length and hue are unknown, and the hue is predicted under conditions where the run length and gradation are known. The reason is as follows. By predicting color elements by dividing them into gradation and hue, if the predicted value and actual value of only one of these elements is not the same, it is only necessary to encode the element whose predicted value and actual value are not the same, which reduces the amount of code. There is an effect that it is possible to reduce. Furthermore, since the gradation is known for hue prediction, the probability of correct prediction is further increased, and the amount of code can be reduced. Also, MH used especially in horizontal mode
In a run length code such as a code, the statistical distribution of the run length differs depending on the gradation, and for codes with a short run length, it is better to use different codes depending on the gradation to improve the encoding efficiency.

However, when the run length is encoded on the gradation side, even if the code word is the same, the run length represented by the code word differs depending on the gradation, so in order to correctly decode the run length code word, it is necessary to The gradation of the long codeword must be known. That is, the gradation must be determined before the run length. Therefore, in the horizontal mode, gradation prediction is performed under conditions where the run length and hue are unknown.

An example of the prediction method will be explained for a two-tone multi-hue image. Prediction of gradation in a run encoded in the vertical mode can be performed by a prediction method similar to the color prediction method explained in FIG. 11, so detailed description thereof will be omitted here. Hue prediction is performed under the condition that the gradation is known, as explained above. FIG. 18 is a diagram for explaining an example of a hue prediction method for a run encoded in vertical mode, and shows the run P whose hue is to be predicted and the run P in its vicinity that is referred to for prediction. FIG. 3 is a diagram showing a relative positional relationship with a run. In this example as well, the run group (2) is limited to runs within one scanning line immediately above the run. Also the 13th
Figures (a) to (c) show cases where the gradation YL of the run L to the left of the run P is equal to the gradation YP of the run P (therefore, the hue HP is necessarily different from the hue HL), and 18th
Figures (d) to (f) show cases where the gradation YL is different from the gradation YP. In addition, in each of the cases shown in FIGS. 18(a) to (f), the run group ■1 referred to when predicting the hue of the run P is a subset of the run group V, and the distance from the change point of the run P is A run group is defined as a run group including pixels from -T3 to the change point of the run R on the right side of the run P to the distance marker T4. FIG. 18(a) shows the case where the only run belonging to run l'f Vt is run 1, and in this case the predicted value H6 is H1.
When ≠HL and Y1=YP, it is assumed that the hue is equal to H1, and when the above conditions are not met, the predicted value HP/ belongs to the run I\U, the gradation is equal to YP, the hue is different from HL, and the predicted value HP/ is the highest in run P. Let it be equal to the hue HK of the nearby run. FIG. 18(b) shows run group V.

This shows the case where there are two runs belonging to run 1 and run 2. In this case, if there is only one run in run nv whose gradation is YP and whose hue is different from HL, the predicted value is the value of that run. If there are two runs with the above conditions, run 1, run 2, and the relevant run P.
Due to the relative positional relationship between
Further, when the change point of run 2 is to the right of the center pixel of the run, the predicted value HQ is equal to the hue H1.

Further, when the gradation of run 1 and run 2 is not equal to gradation YP, it is assumed that the predicted Buddhist color belongs to run NU and the gradation is YP and the hue is different from HL and is equal to the hue of the run closest to the run P. Figure 11 (C) shows a case where there are 8 or more runs belonging to run group ■1, such as runs 1 to 4, and in this case, there is a run whose color is CL among runs B'f Vt, and the right When there is a run whose neighboring run is at gradation YP, assume that the hue of this run whose gradation is YP is equal to the predicted value H'P,
In addition, when there are two or more applicable runs, the hue of the run whose center pixel is closest to the center pixel of run P among the applicable runs is assumed to be equal to the predicted value IIF-; If there is no run, it is assumed that the hue of the run that belongs to the run group U, has a gradation of yp, and has a hue different from HL and is closest to the run P and the predicted value Hρ are equal. Figure 18(d) shows Run! ! This shows the case where the only run belonging to ¥vl is run 1. In this case, when Y0≠YL and Yl=YP, the predicted value group is assumed to be equal to the hue H1, and when other conditions are other than the above, the predicted value group is the run group U.
It is assumed that the gradation belongs to yp and is equal to the hue of the run closest to run P. FIG. 18(e) shows a case where there are two runs belonging to the run group v1, run 1 and run 2. At this time, when there is only one run in the run group vl whose gradation is YP, the predicted value H + horizontal gradation Y2 is equal to the hue of the run, and when the gradations of run 1 and run 2 are both equal to the gradation YP, run 1, run 2, and The predicted value H6 is determined based on the relative positional relationship of the run P. Further, when the gradations of run 1 and run 2 are both not equal to gradation YP, it is assumed that the run belongs to run group U and has a gradation of YP, and the hue of the run closest to run P is equal to the predicted value H6. FIG. 18(f) shows a case where there are eight or more runs belonging to run group V1, such as run 1 to run 4, and in this case, the predicted value group is seek. The above description of a method for predicting color element by element is a description of a method for predicting gradation and hue of a run encoded in vertical mode.

Next, a method for predicting the gradation and hue of a run encoded in horizontal mode will be described. Gradation prediction is performed under conditions where the run length and hue are unknown, as described above.

FIG. 14 is a diagram for explaining the gradation prediction method,
FIG. 3 is a diagram showing a relative positional relationship between a run P whose gradation is to be predicted and a run in its vicinity that is referred to for prediction. In this example as well, the run group V is limited to runs included in one scanning line immediately above the scanning line including the run P. Also, the run m V+ to be referred to when predicting the gradation of the run P
Let be a subset of the run l\V and a run group including pixels located from a distance n'rs to a distance 10T6 from the change point of the run P. FIG. 14(a) shows a case where only run 1 belongs to run group V1, and in this case, the predicted value Y
6 is equal to gradation Y1 when c1≠cL, and c1=
Since we are considering a two-tone image at cL, it is assumed to be equal to the other gradation different from the gradation AYL. FIG. 14(b) shows a case where there are two runs belonging to run group Vl, run 1 and run 2, and in this case, H1=HL and H2≠HL.
When H1 and H2=HL, the predicted value Y6 is assumed to be equal to the tone Y2, and when H2=HL and H,≠HL, the predicted value YP' is assumed to be equal to the tone Y1, and when H0≠
When HL and H2≠HL, if the change point of run 2 is to the left of the change point of the run P due to the relative positional relationship between run 1, run 2, and the run P, the predicted value Y6 is equal to the gradation Y2. Further, when the change point of run 2 is to the right of the change point of run P, the predicted value Y6 is equal to the gradation Y1. FIG. 14(c) shows a case where there are eight or more runs belonging to run group V1, such as run 1 to run 4. In this case, if there is one or more runs with color cL in run group V1, this color It is assumed that among the runs to the right of the run cL, the gradation of the run whose change point is closest to the change point of the run P is equal to the predicted value Y/. Further, when there is no corresponding run, the predicted value Y6 is assumed to be equal to the other gradation level different from the gradation level YL.

The above explanation in FIG. 14 is an explanation of a method for predicting the gradation of a run encoded in horizontal mode. As explained above, prediction of the hue of a run encoded in the horizontal mode is performed under conditions where the gradation and run length are known. This condition is the same as the winter crop when predicting the hue of orchids encoded in vertical mode, so the method for predicting the hue of orchids encoded in horizontal mode is as follows: This method is similar to the method for predicting the hue of a run.

The types of codes used in the two-tone multi-hue image described above are, for example, the codes shown in FIGS. 7 and 10 above, and instead of the color identification code (ID), the hue identification code and Two types of codes are required.

Regarding the gradation, since there are two gradations, it is only necessary to indicate that the predicted value and the actual value are different. In the horizontal mode, a code indicating hue and gradation may be used immediately after the code indicating horizontal mode, and the code indicating gradation is the run length of the run length code.
0'' code may be substituted.The type and code word of these codes should be determined according to the statistical properties of the color image, and any code that does not lose its uniqueness during decoding may be used. A code may also be used.

The above explanation was about the prediction method for images with two gradations, but prediction can also be performed using the same method for color images with eight or more gradations. Since this is larger than two gradations, this may be used.

FIG. 15 is a diagram showing an example of using the correlation of gradations, and shows a case where the gradation of the run P is predicted under conditions where the hue and run length are unknown. For example, H, = H2 = HL. In some cases, prediction may be made such that the predicted value Y6 is equal to yt, +y, -y. Further, when the predicted value and the actual value are different, the actual value may be encoded, or the difference between the predicted value and the actual value may be encoded. When the difference between the predicted value and the actual value is encoded in this manner, the amount of code may be reduced because the value of the difference between the predicted value and the actual value has a distribution concentrated around 0. Furthermore, in the above description of the method of encoding color elements by dividing them into gradation and hue, the same effect can be obtained even if the order of prediction of gradation and hue is exchanged.

It goes without saying that the run color prediction method described above is not limited to the above example.

For example, FIG. 11(C), FIG. 18(c), and (f).

In each case such as in FIG. 14(C), if there are multiple candidate runs for which the color or color element value of that run can be adopted as the predicted value of the color or color element value of run P, the above explanation will be applied. In this case, it is assumed that the run whose center pixel is closest to the center pixel of run P is selected, but this can also be done by selecting the leftmost run or the rightmost run, or by choosing the run that has the closest pixel to the center of run P. Almost the same effect can be obtained even if a run is selected in which the pixels of are adjacent to each other. In addition, as a condition for it to be a candidate run, the color or color element value of the immediately preceding run must be equal to the color or color element value of run L, but in addition, the color or color of the previous run before the candidate run It is also possible to set the condition that the element value of is equal to the color or color element value of the run immediately before run L, and to set the condition that the run length of the candidate run is not much different from the run length of run P. It is possible.

In addition, in each case in the explanation of the above prediction method, the run group V
Distance W & from the relevant run P when determining the run belonging to 1
The Tl to T6 values should be appropriately selected according to the statistical properties of the color image and are inherently variable. Furthermore, the 11th
18 and 14, how to classify the relative positional relationship between the predicted run and the neighboring runs referred to for prediction, and the actual value of the color of the reference run in each classification. The method for associating predicted values should be changed according to the type of statistical properties of the color image. In addition, in the examples explained above, the run group V is a run that is included in one scanning line directly above the scanning line that includes the relevant run. Although the prediction method will become more complex if the prediction method is expanded to include multiple runs, it is expected that the probability of making correct predictions will further improve.

FIG. 16 is a diagram illustrating an example of the effect when the run group V is expanded to two upper scanning lines. This figure is a diagram showing the relative positions of runs that are referred to in order to predict hue in a two-tone multi-hue image, assuming that the gradation and run length of the run P are known. In the method described with reference to FIG. 18, the hue predicted value H/ is assumed to be equal to the hue HA of run A because the change point of run B is on the right side of the central pixel of run P.

However, when referring to the colors of run C and run D, it is predicted that runs of the same color are connected downward to the left from run C to run A and from run D to run B to form an area. It can be predicted that the predicted hue value H6 is equal to the hue H of run B. In the simple color images targeted for encoding in this invention, pixels of the same color often constitute a two-dimensional area, so this prediction method can increase the probability of correct prediction. It is.

In addition, in the example of the color prediction method explained so far, if the run P for which the color or its element is to be predicted is included in the scan line at the left end of the color image, or if the run P is included in the left end scan line of the color image, The run group V or the run group U that should be referred to at the end of the prediction if it is located at or near the right edge; it may happen that it does not exist. To deal with this, assume an area with a color equal to +C; You should leave it there).

Furthermore, regarding the left and right ends of the color image, it is assumed that the right end of the single upper scanning line and the left end of the lower scanning line are connected, and the run group V and the run 8￥U
By the way, the encoding method explained above is an encoding method that encodes the change point of a run and the color of the run independently, but if the change point of the run and the color of the run are related An example of a method of encoding by associating the changing point of a run with the color of the run will be briefly explained below. The symbol used is the same as the symbol used in the above explanation.Special code for encoding change point a1. tac
When p=c and CR=C2, it is encoded in vertical mode, and when the above conditions are not met, it is encoded in horizontal mode. There are four ways to encode the right and left boundaries of the run P as shown in the table below. Among these four types of encoding, the color of the run P is determined by the above conditions for vertical mode. In addition, when both the left boundary and the right boundary of the run P are encoded in the horizontal mode, the predicted values can be obtained by the method explained in FIGS. 11, 1a, and 14. I can do it,
Furthermore, since the conditions that CP=C and CR=Q are not satisfied, the probability of correct prediction may increase. Furthermore, since there is no need to predict the color of a run encoded in the vertical mode, the color prediction circuit can be simplified. The pass mode is used as appropriate as explained in FIGS. 5 and 8.

In this explanation as well, when handling color, it is also possible to encode it by dividing it into two elements, gradation and hue. In the above explanation as well, the method for classifying vertical mode and horizontal mode and the method for predicting color should be determined based on the statistical properties of the image.

Note that in all the examples of encoding run positions so far, we have explained that the changing point of the run, that is, the leftmost boundary, is encoded, but the rightmost boundary of the run can also be encoded using the same method. Needless to say, it can be modified and the same effect can be obtained. Furthermore, in the explanation of the block diagram of the embodiment shown in FIG. 8 regarding the configuration of the color image encoding and decoding circuit, the memory (4) and the memory @ are used as memories for storing change point addresses and colors. Another embodiment of the color image encoding and decoding circuit having the same effect as a memory for storing image data pixel by pixel is obtained. However, at that time, the input to the memory (4) becomes the image data input to the run detection circuit (3), and the input to the memory α becomes the pixel-by-pixel image data output from the image data restoration circuit α. In addition, in the explanation so far, the position specification method selection circuit (5), the two-color prediction circuit (6), the two-position code decoding circuit (9), and the color prediction circuit θ0 may be configured by a memory logic circuit such as a microprocessor or the like. It has the effect of processing.

As explained above, in the color image encoding method according to the present invention, each run in a scanning line is coded with a code representing the color of the run and a code representing the position of the pixel at the left or right boundary of the run within the scanning line. When encoding a color image in the order of scanning lines by converting to , nU of the runs before the run in the scan line including the run to be converted and the runs in the scan lines before the scan line including the run to be converted. With reference to group V, the color of the orchid is predicted from the run group U and the run group ■ using a predetermined method, and if the prediction is correct, the code representing the orchid color is omitted, Regarding the run boundary position, if there is a run boundary position of a run belonging to run group V within a predetermined range in the left and right direction from the run boundary position of the run, the run boundary position of the run belonging to run group V mentioned in 1. Converts to a code that represents the relative position from the position, and in other cases, from the run boundary position closest to the left of the run boundary position of the run among the run boundary positions of runs belonging to run group U or run group V. Since the data is converted to a code representing relative position, the amount of data for the code representing the run color and the code representing the run boundary position is significantly reduced, allowing color images to be encoded efficiently, and there are no limitations on the image representation ability. This has a great practical effect in that it does not occur.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing an example of encoding in a conventional color image encoding method, FIG. 2 is a diagram showing the principle of encoding in a color image encoding method according to the present invention, and FIG. 8 is a diagram showing a color image according to the present invention. FIG. 4 is a block diagram showing an example of the device configuration of the encoding method, FIG. 4 is a diagram for explaining the relative positional relationship of run boundaries, and FIG. 5 is a diagram for explaining an example of a method for encoding run boundary positions. 6 is a diagram showing an example of a general flowchart of encoding run boundary positions, FIG. 7 is a diagram showing specific examples of code types, and FIG. 8 is a diagram showing a method of encoding run boundary positions. FIG. 9 is a diagram for explaining another example; FIG. 9 is a diagram showing an example of a schematic flow chart of encoding such a run boundary position; FIG. 10 is a diagram showing a specific example of code word types; The figure is a diagram for explaining an example of a run color prediction method, FIG.
FIG. 8 is a diagram for explaining an example of a method for predicting the hue of a run in vertical mode, FIG. 14 is a diagram for explaining an example of a method for predicting gradation of a run in horizontal mode, and FIG. FIG. 16 is a diagram illustrating an example of a method for predicting run gradations in a multi-gradation image. FIG. 16 is a diagram illustrating an example of the effect of referring to multiple upper scanning lines in predicting run colors. , FIG. 17 is a diagram for explaining an example of a method of encoding run boundary positions in association with run colors. (1)...Color image encoding circuit, (2)...Color image decoding circuit, (3)...Run detection circuit, (4),
(1, a...Memory, (5)...Position encoding method specification circuit, (6), DI...Color prediction circuit, (7)...
Encoding circuit, (8)... Code identification circuit, (9)...
Position code decoding circuit, 01)...Color code decoding circuit,
03...Image data restoration circuit representative Shin Kuzuno - Ri N
()()
−′ ＼
Fig. 9 Fig. 11 Fig. 12 Fig. 13 Fig. 1-IK414L YP=YL =YK HK bowl 1-IL YP2 YL='l') Figure 15 Figure 16 Figure 17

Claims (1)

[Claims] Within a scan line, runs of the same color are represented by a code representing the same color (i.e., run color) and a code representing the position of the left or right end pixel of that run within the scan line (i.e., run boundary position). In the case where a color image is encoded in constant distortion line order by conversion, a group U of runs before the run in the scan line including the run to be converted, and a group U of runs before the scan line including the run to be converted. With reference to the run group V, the run color is predicted by a predetermined method from the run group U and the run group V, and if the prediction is correct, the code representing the run color is omitted, For run boundary positions,
If there is a run boundary position belonging to the run group V within a predetermined range in the left and right direction from the run boundary position of the run,
It is converted into a code representing the relative position from the run boundary position of the run belonging to the run lv above, and if the run boundary position does not exist within the predetermined range, the run boundary position of the run belonging to run IU or run group ■ is converted. A color image encoding method characterized in that the run boundary position of the run is converted into a code representing a relative position from the nearest boundary position before the run boundary position.