JPH0334680A - Picture processor - Google Patents

Abstract

PURPOSE:To eliminate dots in a low-density part and a link of void in a high- density part in the error diffusion method and a chain texture in middle and low-density parts by providing an already binarized window in the periphery of a weighted picture element and changing the binarization threshold of the noticed picture element based on the dot pattern in the window. CONSTITUTION:A window means 10 which demarcates the area of an already binarized signal in the periphery of the noticed picture element, a threshold determining means which determines the threshold for binarization of the noticed picture element in accordance with the window means 10 and the density level of the noticed picture element of an inputted multilevel picture signal, and a binarizing means 30 which binarizes the weighted picture element by the error diffusion method in accordance with this threshold are provided. A pattern, especially, a link of dots in an oblique direction in the window is detected, and the density level of the original signal of the weighted picture element is discriminated to recognize the chain of the texture, and the threshold of the weighted picture element is changed in accordance with the texture. Thus, dots and a link of void as well as an uncomfortable chain texture in middle and low-density parts are eliminated.

Description

[Detailed description of the invention]

[Industrial Field of Application] The present invention relates to an image processing device for converting multi-value image information into multi-value image information having a smaller amount than the multi-value image information, or into binary image information. [Conventional technology 1 Conventionally, colorants of one color or multiple colors have been used,
Furthermore, there are printers that can print out in binary or 3-4 values by means of modulating the dot diameter. On the other hand, the number of image information input from an image input device such as an image scanner is generally greater than the number of values of image information that can be output from this printer. For this reason, conventionally, multivalued image information from this input device must be converted into information with a number of values that can be output by a printer or the like. That is, input multivalued image information from an input device (for example, one color, several pits per pixel, about several tens of bits),
It must be converted into binary or 3-4 value image information that can be output by a printer. Conventionally, as a general method of converting image information for printing out from such a printer, for example, the means of expressing halftones generally include a dither method, a density pattern method,
A conversion method such as error diffusion method should be adopted.

[Problem to be solved by the invention]

However, the conventional conversion method has the following problems. That is, in the dither method, resolution and gradation are contradictory conditions, and if high gradation is to be obtained, a larger dither matrix must be constructed. Furthermore, in the density pattern method, one pixel of original information is divided into a plurality of pixels, resulting in a decrease in resolution. Therefore, an error diffusion method has been proposed as a binarization method that satisfies both resolution and gradation to some extent. but,
The disadvantages of this method include the connection of dots and white spots, and the unpleasant chain-like texture that appears regularly, which significantly deteriorates the image quality. For example, when considering an output device such as an inkjet recording device that holds two types of coloring materials, a dark ink and a light ink, for each color, that is, can output three values per color, differences in ink color, etc. Therefore, the conventional three-level error diffusion method has a problem in that tone jumps occur in a certain density variation area, resulting in false contours. [Means for Solving the Problems] The present invention has been made with the aim of solving the above-mentioned problems, and includes the following configuration as one means for solving the above-mentioned problems. That is, in an image processing conversion device that converts multivalued image information into a binary signal using the error diffusion method, there are already two
A digitized window is provided, and a threshold value changing means is provided for detecting a dot pattern within the window and changing the binarization threshold value of the pixel of interest based on the detected pattern. In addition, in an image processing conversion device using an error diffusion method that converts multivalued image information into multivalued image information that is less than the multivalued image information, the pixel of interest of the converted signal that has already been converted into fewer pixel signals A window is provided in the vicinity, and a threshold value changing means is provided for detecting a dot pattern within the window and changing the threshold value of the pixel of interest based on the error diffusion method based on the detected dot pattern. [Operation] In the above configuration, when multivalued image information is binarized by the error diffusion method, a window of surrounding pixels that have already been binarized is provided, and the pattern within the window, especially the connection of diagonal dots, is By detecting the texture and determining the density level of the original signal of the pixel of interest, the chain of texture is recognized, and the threshold value of the pixel of interest is changed according to the texture. Connections between dots and white spots in the error diffusion method and unpleasant chain-like textures in medium and low density areas can be eliminated at low cost and with a simple configuration. Furthermore, by detecting whether the dot connections are textures or L or the original signal of the original image, good and appropriate binarization can be performed without losing the original image information. Furthermore, multivalued image information (several pittles, 10 bits/pixel)
In the halftone expression means that converts the signal into 3 or 4 values per pixel using the multilevel error diffusion method, a window is provided in the converted signal and the connection of dots within the window is detected. By recognizing the texture and changing the threshold of the pixel of interest according to the texture,
It is now possible to eliminate the dots and connections between white spots and unpleasant chain-like textures in medium and low-density areas using a simple and inexpensive method, which have traditionally been very problematic in the multilevel error diffusion method. can. In addition, so-called multi-value printers can detect and recognize the switching part of a collection of equal-level dots in a window, and convert the threshold value of the pixel of interest in the same way to output approximately 3-value or 4-value data. Problematic false contours can also be reduced by switching the coloring material in the output. Further, by determining whether the texture in the window is an original signal or one generated by halftone processing, good and differential halftone processing can be realized without losing the original image information. [Example] Hereinafter, an example according to the present invention will be described in detail with reference to the drawings. [First Embodiment 1 FIG. 1 is a diagram showing the functional configuration of an embodiment according to the present invention. The image processing conversion device of this embodiment converts a multi-valued image signal into a binary image signal and outputs it. It is an image processing transformation device. In the figure, a dashed-dotted line block 100 indicates an image processing conversion device that binarizes a multivalued image signal input from the outside and outputs it as a binary image signal. The image processing conversion device 100 shown in FIG. 1 shows only the binarization control functional portion of all image processing means, and other necessary components are omitted from the figure. This image processing conversion device 100 is configured as, for example, a host computer section (as hardware or software).
It is possible to incorporate it into an output device such as a computer or a printer, or an input device such as an image scanner. In the image processing and modification device 100, 10 is a window means for performing area segmentation of the already binarized signal around the pixel of interest, and 20 is the window means 10 and the input multivalued image signal is divided into regions according to the density level of the pixel of interest of the input multivalued image signal. Threshold determining means 30 determines a threshold for binarizing a pixel, and numeral 30 is a binarizing means that performs a binarizing process on the pixel of interest using an error diffusion method in accordance with the threshold determined by the threshold determining means 20. The binary signal binarized by the binarization means 30 is also output to the window means 10, and is used for subsequent binarization of the multi-valued image signal. FIG. 2 shows a block configuration for achieving the functions shown in FIG. 1 of this embodiment. In FIG. 2, 1 is stored in the ROM 7, for example, the third
A central processing unit (CPU) 2 controls the entire apparatus of this embodiment according to the control procedure shown in the figure, and Ilo 2 controls an interface with a connected external device 8 that outputs a multivalued image signal. A multivalued image signal from the external device 8 is input through this l102. 3 is an in-window patterning circuit that performs patterning processing on the binarized signal for a window in the sub-scanning direction; 4 is a threshold determining circuit that determines a threshold for binarization; and 5 is a threshold determining circuit 4. An error diffusion circuit performs binarization processing using an error diffusion method according to a determined threshold value, 6 is a RAM, and 7 is a ROM that stores the above-mentioned programs and the like. Further, 8 is an external device that outputs a multivalued image signal and the like. Hereinafter, the image processing of this embodiment having the above configuration will be explained in the third section.
This will be explained with reference to the flowchart shown in the figure. The apparatus of this embodiment binarizes the first few lines of the multivalued image signal to be binarized inputted from the external device 8 up to 0 using the normal error diffusion method. This is because if there is no already binarized signal, a window for binarization processing cannot be applied, which is unique to this embodiment. When the binarization process using the normal error diffusion method for the lines used for the window in the sub-scanning direction is completed, the process shifts to the process shown in FIG. 3, and first, step S1
This binary signal is converted into a window pattern circuit 3.
and reads the pattern of data in the window. An example of the window configuration in this case is shown in FIG. In Fig. 4(A), the "*" mark indicates the pixel of interest, and the windows a to h surrounded by square cells are the windows of the binarized signal that have already been binarized, and the windows not enclosed by squares and indicated by numbers are is a diffusion matrix showing commonly used error distribution using the error diffusion method. Furthermore, the arrows in the figure indicate the binarization scanning direction. In the window shown in Figure 4(A), 8
It is configured to take in two binary signals. and,
By arranging this binarized data as data a to h as shown in FIG. 4(B), the window data can be treated as 1-byte data from now on. Preferably, this window is configured so that diagonal connections of dots can be detected. In other words, in binarization processing using the error diffusion method between shells, dots in highlighted areas and connections between white spots often occur in diagonal directions, resulting in an unpleasant chain-like texture in medium and low density areas. This is because dots are lined up diagonally to create a regular pattern. No. 13, which has already undergone binarization processing, detects whether or not dots are lined up diagonally within the window, and feeds this back to the threshold value of the pixel of interest to prevent further diagonal dots from connecting. Should be. For 1-byte data as shown in FIG. 4(B), 256 patterns may occur, but it is better to change the threshold only when there are several patterns in which diagonal dots are connected. In the subsequent step S2, the level of the original signal of the pixel of interest (before addition of errors of adjacent pixels) is determined. Here, even if dots are connected diagonally, it is determined at the same time whether the original signal is originally image information of a diagonal line or whether the dots are connected diagonally by chance using the error diffusion method. That is, the density level of the original signal of the pixel of interest is a high density level (for example, if the input image signal is an 8-bit multivalued signal, "0" is the white level, r
If 255J is the black level, it is determined whether the black level is at r200J or higher. If it is at a high density level, whether the original signal is originally a diagonal line image signal or
Or, it can be determined that there is a strong possibility that it is an edge part, and if the density level is not high, it can be considered as a diagonal line due to the connection of dots in the highlight part, or as part of a chain of a chain-like texture in the medium or low density part. can. Therefore, the next step S
If 3 is outside the high concentration level, the diagonal dots will connect,
In order to break the chain, a threshold value is determined in the threshold value determination circuit 4. Thereafter, in step S4, errors of adjacent pixels are added, and in step S5, the error diffusion circuit 5 performs binarization processing. Then, in step S6, it is determined whether all the signals have ended,
If all the signals have not been completed, the process advances to step S7, the pixel of interest is scanned by one pixel, and the process returns to step S1. Thereafter, the same operation procedure is repeated to perform binarization processing on the next pixel of interest. When the binarization process for all pixels is completed, this process ends. Here, errors diffused by the line buffer for several lines, binary signals temporarily stored for binarization and window processing, etc. are temporarily stored in a predetermined area of the RAM 6. In addition to the program, the ROM 7 also stores diffusion matrices, etc., and a table for determining the threshold value according to the pattern in the window and the density level of the pixel of interest is stored and saved in advance in this ROM7. It is more efficient to make decisions by referring to this table. FIG. 5 shows an example in which the threshold value is changed in the window of this embodiment shown in FIG. 4. In the example of FIG. 5(A), only C1j in FIG. 4(A) is "l".
(black), and the other dots in the window are "O" (white), ie, the window data is (24). Assuming that the original signal of the pixel of interest is 80 (8-bit multivalue, rOJ is the white level and r255J is the black level), there is a high possibility that it is part of a chain due to a ridge-like texture in the medium density region. For this purpose, the threshold value of one large pixel is increased to add randomness to the dots. The example in FIG. 5(B) is shown in FIG. 4(A). e, g are "l" (black) and other dots in the window are "0" (white), that is, the window data is (5A
) )l. If the original signal of the pixel of interest is lOO, then
Similar to the example in FIG. 5(A), this is considered to be inside the texture of the chain. In other words, if the pixel of interest is "O" (white),
A positive error is propagated to the pixel to the right of the pixel of interest, and the pixel to the right tends to become rlJ (black), which may lead to a chain. Therefore, in this embodiment, the threshold value of the pixel of interest is lowered to make it easier for the pixel of interest to become "1" (black). unknown). The example in FIG. 5(C) is (24)H, which is the same as the window data in FIG. 5(A). However, since the original signal of the pixel of interest is 220, it is unlikely that chains occur in the middle and low density regions, and in this case, the threshold remains at the middle level (127) as in the normal error diffusion method. Fix it. As mentioned above, the threshold value is determined based on the 1-byte data in the window and the original signal of the pixel of interest, and the threshold value is divided into several levels such as high (180), middle (127), and low (70). It is also possible to have a configuration that is divided into . Further, as described above, a method may be adopted in which the correspondence between 1-byte data of the window and the threshold value is stored in advance in the ROMT as a table. [Second Embodiment] In place of the window configuration described above, the window configuration shown in FIG. 6 can provide similar effects and is included in the scope of the present invention. FIG. 6(A) is an example in which a window is provided with eight pieces of data from a to h, which are replaced with the same 1-byte data as in FIG. 4(A). Compared to the window in FIG. 4(A), the window in FIG. 6(A) also takes into account the information above and to the left of the pixel of interest.
This is more effective for recognizing surrounding textures. However, there is no difference in the connection of dots in the low-density area compared to the window in Figure 4 (A), which incorporates a lot of information in the diagonal direction.
Alternatively, the method shown in FIG. 4(A) may be more suitable, or a combination of both methods may be used. FIG. 6(B) is an example in which the window size is made larger so that texture chains and dot connections can be more clearly recognized. The embodiment of this window is configured to create 2-byte data from 15 pieces of data from a to 0. and,
The comparator 61 detects that the original signal of the pixel of interest is at a high density level (for example, an 8-bit multivalued image signal of 200 or more (however, r2
55J is the black level and "0" is the white level)), and if it is a high density level, it is "O", and if it is not, it is rl.
J, and this may be placed in the MSB of the upper byte as a flag. This configuration is more efficient when creating a table corresponding to threshold values in the ROM 7. This flag system can be configured similarly even in a 1-byte data window by setting the window size to 7 pieces of data. [Third Embodiment] FIG. 7 shows an example of a third embodiment of the present invention in which random numbers are applied to the threshold value determining means. In the third embodiment shown in FIG. 7, if the threshold value of the pixel of interest is increased or decreased by a constant value according to texture recognition, there is a possibility that another texture will occur due to this regularity. This was done in view of the fact that the texture noise is significantly reduced compared to the unexampled example. In the third embodiment, when the threshold value of the pixel of interest is increased or decreased by a constant value in accordance with texture recognition, a random number is used to change the threshold value to add some randomness. In FIG. 7, the threshold level is rxJ, and the threshold is classified into three states: high state (when you want to raise it), middle state, and low state (when you want to lower it). Then, let the probabilities of the threshold values be fl (x) f, (
x) fz (x). Here, f, (x) is a function between 20OA-127 centered on the 170 level, and f-(x) is a function between 50-127 centered on the 80 level.
The function between fs(x) is 100 to 1 centered at 127
Let it be a function between 50 and 50. Assuming that the threshold value is determined by random numbers, the probability of occurrence of random numbers is expressed as fl (x), t2 (x) + f'3 (
If the threshold value is set as shown in the curve x), for example, when the threshold value is increased, there is a possibility that the threshold value is not fixed at 170 but is determined to be around 170. Therefore, here:
Randomness will occur. Also, since the probability of random number generation is changed, f, (x
), the probability of reaching 170 is highest, and the probability of occurrence decreases as the distance from 170 increases. Therefore, random noise is less likely to occur, and sandy images can be avoided. Furthermore, in this embodiment, f-(x) is also provided as a function centered around the threshold value 127, but due to cost considerations, f and (x) are fixed at 127 (i.e., delta function δ(x-12)=1). It's good as well. In this way, according to the third embodiment, texture and dot connections can be prevented by giving the threshold some randomness.

[Fourth Embodiment] In addition, in the texture described above, by counting the number of occurrences of texture (particularly a connection of diagonal dots) and using the counted number as a means of determining a threshold value, an even better image can be obtained. Can be processed. FIG. 8 shows a block diagram of a fourth embodiment of the present invention in which the number of occurrences of this texture is counted and the counted number is also used as a means for determining a threshold value. In FIG. 8, the same components as those in FIG. 2 described above are given the same reference numerals, and detailed explanations will be omitted. FIG. 8 shows a configuration in which a counter 9 is added as an additional function to the embodiment shown in FIG. That is, in this embodiment, when a certain texture occurs, the number of occurrences is counted, and therefore, as the count increases, the noise of the chain-like texture increases. Therefore, as the number of counts increases,
Increase the amount of up/down of the threshold value to make it possible to prevent the dots from connecting. By performing such processing, the window of the binarized signal by the counter 9 becomes equal to a wide equilaterally curved window, and more accurate texture recognition and accompanying threshold changes can be performed at low cost. [Fifth Embodiment] FIG. 9 shows control in a case where the processing shown in FIG. 3 described above is configured to be faster. The processing in FIG. 9 shows a fifth embodiment of the present invention, and realizes faster processing than the flowchart shown in FIG. In the fifth embodiment, first, in step Sll, it is determined whether the original signal of the pixel of interest is a high density portion (whether it is 200 or more). Then, if it is a high concentration area, the process advances to step S15, and a normal error diffusion method is performed while keeping the threshold value fixed. Only when it is not a high-density portion does the process proceed to step S12, where window processing unique to this embodiment is performed. First, in step S12, a window pattern reading process similar to step S1 in FIG. 3 is performed, and in the subsequent step S13, the texture pattern within the window is determined. If there are no diagonally continuous dots within the window, the process proceeds to step 515, and the error diffusion method is performed with the fixed threshold value unchanged. On the other hand, if there is a succession of diagonal dots, that is, a texture bud, then in step S14, the same threshold value changing process as in steps S2 and S3 in FIG. 3 is performed, and the process proceeds to step S15. In step S15, error addition processing of adjacent pixels is performed as in step S4 in FIG. 3, and in subsequent step S16, binarization processing is performed using the error diffusion method as in step S5 in FIG. 3. Then, in step S17, as in step S6 in FIG. 3, it is checked whether the binarization process for all pixels has been completed. If not, in step S18, a new pixel of interest extraction process similar to step S7 in FIG. 3 is performed. Return to step Sll. In the operating procedure of this embodiment described above, the range of the processing area is smaller than the processing shown in FIG. 3 described above, and higher speed processing can be realized. Also, assuming that we only handle textures in medium density areas,
For example, the judgment of step Sll is (70 × original signal < 120
) etc., the speed can be further increased. As explained above, in the error diffusion method, problems such as dots in low-density areas, connections between white spots in high-density areas, and unpleasant chain-like textures in medium and low-density areas have traditionally been very problematic. can be eliminated at low cost and with a simple configuration. Furthermore, by detecting whether the dot connections are textures or original signals of the original image, good and appropriate binarization can be performed without losing the original image information. [Sixth Embodiment] In the first to fifth embodiments described above, processing for converting multivalued image information into binary image information has been described, but the present invention is not limited to the above examples. Needless to say, the present invention can be applied to processing for converting multivalued image information into multivalued image information having a value equal to or less than the input value. Hereinafter, an example will be described in which the present invention is used to convert a human image signal into input multi-value image information with a lower multi-value number. Each of the following embodiments can also be realized, for example, by the image processing conversion device having the hardware configuration shown in FIG. 2 described above. The following description will be made based on a wiring diagram schematically showing image processing using the above-described block configuration. FIG. 10 is a block diagram of main parts of the image processing conversion device of this embodiment. The image processing and modification device of this embodiment shown in FIG. 10 shows the halftone expression control function portion of all the image processing means, and other components are omitted from the figure. For example, this part can be installed inside a host computer (as hardware or software) or an output device such as a printer.
Alternatively, it can be incorporated inside an input device such as an image scanner. In this embodiment, an example in which an input multilevel signal is ternarized will be described as halftone expression. First, a multivalued image signal (several bits of Lumi or several bits) 1.
, enter. This input signal 1 my is sent to input correction means 101 and multivalue conversion (ternary conversion) means 102 . The input correction means 3.01 performs input correction on the pixel of interest by adding errors caused by surrounding adjacent pixels in a process similar to the normal error diffusion method, and outputs a correction signal I'xy. On the other hand, the multi-value conversion means 102 compares the input signal I xy with the multi-value conversion threshold T 1111, converts the input multi-value signal into a three-value signal, outputs it as P', A, and outputs it as a threshold value determination means 109. send to The multilevel conversion means 103 uses the correction signal 1'□ in the input correction unit 101 and the multilevel conversion determination threshold T' in the threshold value determination unit 109.
The input multi-level signal is converted into a ternary signal p xy
Convert to This is true even when the connected output device is a printer that can output up to 3 or 4 values for each color using, for example, dark and light ink, or dot modulation, instead of 2 values for each color. , correspondingly, it is configured to be able to express halftones. For example, when performing three-value output, the multi-value conversion means 103 compares the input signal with a threshold value, converts it into three types of signals (P xy) of (1, station, 0), and the actual signal is (2 ,L,O
), which expresses intermediate tones. Also, a ternary signal P. is also sent to the difference calculation means 104, the difference with the input signal 1゛□ is calculated, and the result is sent to the error distribution calculation means 105 as an error signal El. The error distribution calculation means 105 distributes the weighting of the error signal Exy using the diffusion matrix shown at 106 in FIG. sent to. Further, numeral 107 indicates a window in which a ternary signal that has already been ternarized is stored in a memory and is hung around the pixel of interest (marked with *). In this embodiment, the threshold value T1. The method for determining y is distinctive, and the method for determining the interval value T'my will be described below. In general, ternarization using the error diffusion method has the same problems as binarization, although there are differences in degree. To solve this problem, consider a window pattern as shown in Figure 11 ("l" is the black level, "0" is the white level, and "a" is the gray level). It is preferable to adopt a configuration in which connections in directions can be detected. In the figures, (a) to (d) respectively show examples in which dots and white spots are connected in a diagonal direction. That is, a window is applied to the ternary signal that has already been ternarized, and it is detected whether or not it conforms to the patterns shown in FIGS. 11(a) to 11(d), for example. In a window of 8 pixels as shown in Fig. 11, 3
@ = 6561 patterns may be generated, but in this case as well, as mentioned above, as in the examples (a) to (d), patterns in which dots are connected diagonally are registered in the memory. It is desirable to keep it. The size and shape of the window may be any size and shape as long as the connection of diagonal dots can be detected (depending on memory capacity and processing speed). Then, the pattern recognition means 108 checks whether the data in the window matches the registered pattern of diagonally connected dots and white spots. Thereafter, in the threshold determining means 109, if the pattern recognition means 108 detects a connection between dots and white areas, the threshold is changed so as to break the connection. Detects the subtle connections between dots and white spots that are about to appear as an unpleasant chain-like texture, and adds irregularity. Breaks the subsequent chains and suppresses the occurrence of texture. However, even if the dots are connected diagonally, it is necessary to determine whether the original signal was originally image information of diagonal lines, or whether the dots happened to be connected diagonally as a result of processing using the error diffusion method. Therefore, in Fig. 10, the input signal I. is compared with the threshold value T my to obtain a simple ternary signal (P'
, , )). Input signal■. When is set to 8 bits (00□ (white level) FF, (
black level)), threshold T my as CO,. 408, for example, as shown in Figure 11(a).
In the example above, if a texture occurs within the window, the density level near the pixel of interest is determined by the ratio of the number of 1 (black) data and the number of suspicious (gray) data within the window! 4 (gray) closer than the data. That is, in the case of FIG. 11(a), T: c
oH＞1. , it is determined that it is part of the texture, and T
Xy:C0,4≦IIIF#), it is determined that this is due to image information of diagonal lines from the original signal. Note that the same determination can be made in the cases of FIGS. 11(b) to 11(d). In the error diffusion method, the absolute position of the original signal and the absolute position of the image processed by the error diffusion method may deviate due to accumulation of errors. However, by increasing the window size, it is possible to determine whether a diagonal line (part of a character, etc.) in the original signal is part of the texture. In this case, the threshold value determining means 109 in FIG. Determine the threshold value in the multilevel error diffusion method. That is, only when it is determined from the above-mentioned results that a texture has occurred within the window, the threshold value is converted to break the chain of texture. In the case of ternarization, the threshold value requires two insertions, but the threshold value can be determined in several stages such as high, middle, and low for these two types.
It may be stored in the ROM as an LUT (look-up table) in association with the window pattern.

[Seventh Embodiment] FIG. 12 shows a seventh embodiment of the present invention, and components similar to those of the sixth embodiment shown in FIG. 10 are given the same reference numerals and detailed explanations will be omitted. In the seventh embodiment shown in FIG. 12, the fixed threshold value T xy of the simple ternarization means 102 of the sixth embodiment used in FIG. ti is varied in accordance with the window pattern. As mentioned above, even if the dots are connected diagonally within the window, this may be due to the fact that the original pixel signal is originally image information of diagonal lines, or it is due to the appearance of a texture in which dots are connected diagonally by chance due to the error diffusion method. This is due to the need to determine whether this is the case. In the sixth embodiment, the threshold values T are CO, and 40,
Although we have described an example in which the window is fixed to two types, the present invention is not limited to this fixed example. In order to more accurately determine whether the inside of the window is a diagonal line or a texture, the present invention is not limited to this fixed example. , it is necessary to change the threshold value. That is, to explain this embodiment based on the pattern example shown in FIG. 11, the pattern (a) is a pattern with diagonal lines of l (black) on a background of density of a (gray). Determine whether it is a texture of a medium density range or a texture of a medium density range. In the example shown in FIG. 11, it was possible to identify the information using T, , = CO, but in order to accurately infer that it is image information of a diagonal line, it is necessary to have a threshold value T my in a high concentration region (for example, T
, , = approximately EOn). Further, in the sixth embodiment, the pattern (b) is T xy=4
0. However, it is possible to detect diagonal lines at the O (white) level better by shifting T+ty to the highlight area (for example, about T., = 20.) and then comparing. Furthermore, in the patterns (C) and (d), it can be assumed that special (gray) diagonal lines are included in a black background and a white background, respectively, but a threshold value that narrows the castle (gray) area (for example, , T
-y" A On, 60,), it becomes possible to judge whether the line is a gray diagonal line or not. As described above, in this embodiment, the window pattern of the pattern recognition means 108 In response to this, the threshold determining means 109-A sets T. (setting as LOT), and the multi-value (ternarization) means 102 converts the corresponding ternary signal P.
The configuration is such that it is sent to the threshold determining means 109B as 'o'. Although the structure of this embodiment is somewhat more complicated than that of the sixth embodiment, it is characterized in that it can determine whether it is a diagonal line or a texture with high accuracy.

[Eighth Example] Furthermore, the control example of the eighth example of the present invention, which performs image conversion processing at a higher speed than the above-mentioned sixth and seventh examples, is described as a first example.
Shown in Figure 3. In FIG. 13, the same processing as in the fifth embodiment shown in FIG. 9 described above is performed. That is, first, in step Sl 10, it is determined whether the original signal of the pixel of interest is a high-density portion (for example, I xy≧EO,) or a low-density portion (for example, 1., <201). Here, if the pixel of interest is a high-density area or a low-density area, the process advances to step 5150, and a normal error diffusion method is performed with the threshold value fixed. When the pixel of interest is not in a high-density area or a low-density area (for example, in the case of 20.4≦I xy<E OH), the process proceeds to step 5L20 and subsequent steps to perform window processing unique to this embodiment. First, in step 5120, a reading process similar to step S12 in FIG. 9 is performed, and in the subsequent step 3180, the texture pattern within the window is determined. If the dots are not continuous in the diagonal direction within the window, the process proceeds to step 5150, and the error diffusion method is performed with the fixed threshold value unchanged. On the other hand, if there is a succession of diagonal dots, that is, a texture bud, then the same threshold value changing process as in the above-described embodiment is performed and the process proceeds to step 5150. At step 5150, error addition processing for adjacent pixels is performed, and at subsequent step 5160, ternarization processing using error diffusion is performed. Then, in step 5170, it is checked whether the ternarization process for all pixels has been completed, and if it has not been completed, step 3
At step 180, a new pixel of interest extraction process is performed, and step St
Return to 10. In the operating procedure of this embodiment, the range of the processing area is smaller than in the sixth and seventh embodiments described above, so high-speed processing can be achieved. In addition, when using a so-called multilevel printer that uses multiple coloring materials with different color densities as an output device, it is necessary to adjust the standard only to the density region where the coloring materials switch.
It is also effective to adopt a configuration in which a normal error diffusion method is applied to other density regions. False contours tend to occur when changing coloring materials, so this is useful for preventing them and reducing texture. [Ninth Embodiment] Furthermore, FIG. 14 shows a ninth embodiment of the present invention in which random numbers are used as the threshold determining means in the sixth embodiment shown in FIG. 10. In this embodiment, compared to the sixth embodiment shown in FIG. 10, a random number generator 110 is newly added. In this embodiment, the threshold value (T'□) of the pixel of interest is increased or decreased by a constant value according to texture recognition.
Since there is a possibility that different textures may occur due to this regularity, this was done in order to suppress this, and a random number is used to change the threshold value to add some randomness. For example, change the threshold in 3 stages (high, middle, low)
In the case shown in FIG. 10,
High (T'-=EOM, 60H) and middle (T
'o=COH140,4), row (T',,=AO
,,20,). Then, T'xy is determined using random numbers at each stage of high, middle, and low. That is, for example, in the high state, as shown in FIG. 15(a), the threshold T4. Set y so that the random number generation probability is maximized around EOH, and FFH<T', <C
Random numbers are generated between OH and T' and y are determined. In addition, between % and 0, 7 F H <T'-y< 4
It is 08. Similarly, in the case of middle and low, T'xy is determined so as to be a function with varying probability of random number generation, as shown in FIGS. 15(b) and 15(c). In the case of the error diffusion method that uses random numbers for the threshold, sandy images may occur due to random noise; however, in this example, the probability is changed using a random number generation probability function, so there is no problem even though it only produces some randomness. It can create regularity and not become fissy. [Tenth embodiment] Furthermore, in addition to recognizing the texture within the window,
FIG. 16 shows an example of a window configuration diagram to be corrected according to a tenth embodiment of the present invention in which pseudo contours within a window are also recognized as patterns and the threshold value of the pixel of interest is changed. For example, suppose that patterns such as those shown in FIG. 16 (a) to (C) occur during window processing in this embodiment. In this case, if the connected output device is a multilevel printer, false contours often occur as tone jumps when coloring materials are switched, even when the colors are the same. Therefore, (a) to (C
) (the part where dots of the same level form a cluster and switch) is similar to the above example,
This is stored in an LUT (actually, a variety of patterns other than (a) to (C) are stored), and the suitability of the patterns within the window is checked. If the pixel matches the pattern within the window, the threshold value of the pixel of interest is changed to give irregularity to the connection of the pseudo contours, as in the embodiment described above, and a good image is formed. As explained above, in the multivalued error diffusion method, there are problems such as the connection of dots in low-density areas and missing parts in high-density areas, as well as unpleasant chain-like shapes in medium and low-density areas. Textures can be eliminated at low cost and with a simple configuration. In addition, by detecting whether the dot connections are textures or original signals of the original image, good and appropriate multi-value conversion (3-value conversion, 4-value conversion) can be performed without losing the original image information.
can be done. Additionally, false contours that tend to occur using the multilevel error diffusion method can be similarly reduced. [Effects of the Invention] As explained above, according to the present invention, connections between dots in low-density areas and white spots in high-density areas in the error diffusion method, and unpleasant chain-like textures in medium and low-density areas can be eliminated. cheaply,
It can be eliminated with a simple configuration. In addition, by detecting whether the connections between dots are textures or original signals of the original image, it is possible to perform good and appropriate binarization or higher multivalue conversion without losing the original image information. Ru. Additionally, false contours that tend to occur using the multilevel error diffusion method can be similarly reduced.

[Brief explanation of drawings]

FIG. 1 is a functional block diagram of an embodiment according to the present invention, FIG. 2 is a block configuration diagram of this embodiment, FIG. 3 is a flowchart showing image processing control of this embodiment, and FIG. 4 is a flow chart of this embodiment. 5 is a diagram showing an example of threshold value determination in this embodiment. FIG. 6 is a diagram showing a window configuration in the second embodiment according to the present invention. FIG. 7 is a diagram showing an example of threshold value determination in the present embodiment. FIG. 8 is a block configuration diagram of the fourth embodiment according to the present invention; FIG. 9 is a flowchart showing the image processing procedure of the fifth embodiment according to the present invention; Figure 1O is a block configuration diagram of the sixth embodiment according to the present invention, Figure 11 is a diagram showing an example of a pattern requiring window correction in the sixth embodiment, and Figure 12 is a block diagram of the seventh embodiment according to the present invention. 13 is a flowchart showing the image processing procedure of the 8th embodiment according to the present invention; FIG. 14 is a block diagram of the 9th embodiment according to the present invention; FIG. 15 is a threshold value according to the 9th embodiment FIG. 16 is a diagram showing an example of a pattern requiring window correction according to the tenth embodiment of the present invention. In the figure, 1... central processing unit (CPU), 2...
Ilo, 3... In-window patterning circuit%4...
・Threshold value determination circuit, 5...Error diffusion circuit, 6...RA
M, 7...ROM, 8...External device, 9...Counter, 10...Window means, 20...Threshold value determination means, 30...Binarization means, lOO...Image Processing change device, 101... Input correction means, 102, 103...
・Multi-value conversion (ternary conversion) means, 104... difference calculation means,
lO5... Error distribution calculation means, 106... Diffusion matrix to peripheral adjacent pixels, 107... Window applied near the pixel of interest (marked with *), 108... Pattern recognition means, 109.109-A, 109-B...Threshold value determining means.

Claims (10)

[Claims] (1) In an image processing device that converts multilevel image information into a binary signal using the error diffusion method, a window that has already been binarized is provided around the pixel of interest, and the dot pattern within the window is detected and converted into a detection pattern. An image processing apparatus comprising: a threshold value changing means for changing a binarization threshold value of a pixel of interest based on the threshold value of the pixel of interest. (2) The image processing apparatus according to claim 1, wherein the threshold value changing means adds the original signal level of the pixel of interest in addition to the in-window dot pattern as a condition for changing the binarization threshold of the pixel of interest. (3) When determining the binarization threshold for the pixel of interest, the threshold changing means classifies the random number generation probability of the dot pattern within the window into several functions, and determines the binarization threshold based on the random numbers. The image processing apparatus according to claim 1. (4) The threshold value changing means has a function of counting the number of identical dot patterns within the window, and also refers to the counted number to determine the binarization threshold value.
The image processing device described in Section 1. (5) The threshold value changing means performs binarization processing only on the density range information of a predetermined width of the multivalued image information.
The image processing device described in Section 1. (6) In an image processing conversion device using an error diffusion method that converts multi-value image information into multi-value image information that is less than the multi-value image information, attention is paid to the converted signal that has already been converted into fewer pixel signals. 1. A pixel processing device comprising: a threshold value changing means for providing a window near a pixel, detecting a dot pattern within the window, and changing a threshold value based on an error diffusion method for a pixel of interest based on the detected dot pattern. (7) As a threshold value conversion condition for the pixel of interest of the threshold value changing means,
7. The image processing apparatus according to claim 6, wherein in addition to the in-window dot pattern, a pixel signal obtained by simply multi-valued pixel of interest using a predetermined fixed threshold is also added. (8) The image processing apparatus according to claim 7, wherein the threshold value changing means changes the threshold value for simple multi-value conversion according to the dot pattern within the window. (9) The threshold value changing means is characterized in that when determining the threshold value of the pixel of interest, the pixel is classified into several functions in which random number generation probabilities are changed, and the threshold value for simple multilevel conversion is determined by the random numbers in the classified functions. An image processing apparatus according to claim 6. (10) The image processing apparatus according to claim 6, wherein the threshold value changing means performs multi-value processing only on density range information of a predetermined width of the multi-value image information.