JP4775909B2 - Image processing apparatus, image processing method, program, recording medium, and image forming apparatus - Google Patents

Description

  The present invention relates to an image processing apparatus, an image processing method, a program, a recording medium, and an image forming apparatus that perform optimum halftone processing according to density. For example, the present invention relates to a technique suitable for an electrophotographic printer or copier. .

  When outputting multi-valued input images with a device that has only a small or binary output capability, or for the purpose of reducing the amount of data, output images with a value smaller than the number of multi-valued input images There are various methods using halftone processing technology for converting to. Among them, typical halftone processing techniques include a dither method and an error diffusion method.

  The dither method is a method in which a predetermined threshold matrix is compared with input image data, and dots are output if the threshold is exceeded for each pixel. By using this method, it is possible to easily realize a periodic dot arrangement and to output a uniform image with good graininess. However, since the area where dots are likely to be output depends on the pattern of the threshold matrix, the shape of characters or the like may be lost and the sharpness is inferior. Further, when periodic processing such as dithering is performed in the low density portion, if the size (or density) of the dots that can be output is not small, the period in which the dot pattern appears becomes long. Since human vision is easy to recognize low-frequency patterns, that is, patterns with a long period, and conversely, it is difficult to recognize high-frequency patterns, so when the frequency at which dot patterns appear is low, the dot pattern is conspicuous. As a result, the image quality deteriorates. Therefore, it can be said that periodic processing is unsuitable for processing the low density portion.

  The error diffusion method is a process for outputting a dot if each gradation value of input pixel data exceeds a predetermined threshold value, and dispersing the gradation value as an error if not exceeding. When the error diffusion method is used, gradation values are preserved, and the color reproducibility is slightly better than the dither method. Further, since the threshold matrix is not used, the dot output is uniform, the character shape is not easily broken, and the sharpness is excellent. On the other hand, when processing such as the error diffusion method is performed, the dot arrangement to be output is aperiodic, and therefore, in some areas, dot deviation occurs and graininess is impaired.

  Thus, a technique has been devised that combines a dither method with excellent graininess and an error diffusion method with excellent sharpness. Hereinafter, this combined technique is referred to as error-compensated halftone processing. In this error compensation halftone process, a multi-valued input image is quantized by a quantization process using an error diffusion method. On the other hand, a value corresponding to the pixel position is used as the quantization threshold similarly to the threshold in dither processing.

  As a technique for obtaining a halftone processing result utilizing the advantages of both methods by switching between the error compensation type halftone processing and the error diffusion processing according to the feature of an image, there is Patent Document 1. In this method, an aperiodic process such as error diffusion is performed in a low density region, while a periodic process such as dither is performed in a medium and high density region. As described above, by performing non-periodic processing in low density areas where periodicity causes image quality degradation, the phenomenon of dot arrangement patterns being noticeable can be suppressed. By performing this process, it is possible to obtain an image with a high degree of granularity compared to the non-periodic process.

JP 2001-346041 A JP-A-11-331588

  However, by switching between the low density part and the high density part, a sense of incongruity occurs at the boundary. This is shown in FIG. 17 (a). FIG. 17A shows patch images with similar density values processed by the error diffusion method and the dither method, respectively. Note that FIG. 17B shows the entire error diffusion process, and FIG. 17C shows the entire dither process. Obviously, the former has a sense of incongruity at the switching portion compared to the latter two.

The present invention has been made in view of the above problems,
An object of the present invention is to provide an image processing apparatus, an image processing method, a program, a recording medium, and an image forming apparatus that suppress an uncomfortable feeling that occurs in a certain density range while performing an optimum process for granularity for each density. is there.

The present invention provides a corrected input value calculating means for calculating a corrected input value by adding a predetermined weight to an error value around the target pixel position of image information to be input, and calculating the corrected input value. Threshold matrix determining means for determining a threshold matrix to be applied based on an input value or an input value around the target pixel, and threshold determination for determining a threshold to be applied based on the determined threshold matrix and the target pixel position Output gradation value determining means for determining an output gradation value based on the determined threshold value with respect to the corrected input value of the target pixel position calculated by the corrected input value calculating means, and the output gradation value Error value calculating means for calculating a difference between the output gradation value determined by the determining means and the corrected input value as the error value, and transmitting the calculated error value to the corrected input value calculating means. An image processing apparatus, the plurality of threshold matrix state, and are not the average quantization error becomes substantially constant value in the input tone values each threshold matrix is used, the threshold value matrix determining unit, low Select a threshold matrix that is a dispersed dot arrangement in the density area, select a threshold matrix that is a periodic dot arrangement in the high density area, and intermediate between the low density area and the high density area The most important feature is to select a threshold matrix for dot arrangement .

According to the present invention, the processing is performed so that the dispersed dot arrangement is performed in the low density portion, thereby improving the granularity of the processed image, and the processing is performed so that the periodic dot arrangement is performed in the high density portion. Therefore, it is possible to achieve both the graininess and sharpness of the processed image, and by mixing the dispersed dot arrangement and the periodic dot arrangement in the intermediate density region, the uncomfortable feeling caused by switching between the two can be suppressed. Can improve image quality. Furthermore, by having a plurality of threshold matrices and switching them, the optimum halftone process according to the density can be achieved, so the circuit configuration is simplified and the cost is suppressed.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. In the embodiment, the input image is assumed to be 8 bits, that is, 256 gradations, the gradation value 255 is black, and the gradation value 0 is white.

  An example in which the present invention is applied to an electrophotographic printer will be described. FIG. 1 shows a typical laser printer. An image created by a PC application is converted into bitmap data (hereinafter referred to as image data) by a driver. Further, the image data is subjected to halftone processing so that the printer can output it. The image data after the halftone processing is further subjected to the following predetermined processing. An image is created in the laser printer as follows. First, the light source 1 emits light according to cyan image data. In the photoreceptor 2, static electricity is generated by the photoelectric effect. The revolver 8 is rotated on the photosensitive member 2 charged with static electricity, and charged toner is supplied from the cyan developing unit 3 to form an image. Similarly, by forming an MYK image, a CMYK image is formed on the photoreceptor 2. Finally, the electrostatically charged paper conveyed by the conveying belt 7 is brought into close contact with the photosensitive member 2 to form an image on the paper, and an output image is obtained by fixing.

  Hereinafter, the halftone process of this embodiment will be described. FIG. 2 shows the configuration of halftone processing according to the present invention. The image input unit 101 performs, for example, a color copying machine or the like, performs density correction processing and frequency correction processing on image data read by a scanner, inputs an image converted into an image for each CMYK plate, and inputs an input gradation value. Are sequentially sent to the subsequent processing for each pixel. The threshold matrix determining unit 104 determines an optimal threshold matrix for the pixel in accordance with the output of the image input unit 101. The threshold determining unit 105 determines a threshold according to the output of the threshold matrix determining unit 104 and the output of the image input unit 101. On the other hand, the output of the image input means 101 is sent to the corrected input value calculation means 102. Here, a weighted sum with a quantization error described later is taken and sent to the output gradation value determining means 106. The output tone value determining unit 106 compares the threshold value output from the threshold value determining unit 105 with the corrected input value output from the corrected input value calculating unit 102 to determine the output tone value. Further, the error value calculating means 103 sends a value obtained by subtracting the output gradation value, which is the output of the output gradation value determining means 106, from the corrected input value to the corrected input value calculating means 102 as a quantization error.

  FIG. 3A shows the configuration of the corrected input value calculation unit 102. The corrected input value calculation unit 102 adds the correction value to the output from the image input unit 101 by the adder 204 and outputs the result. The correction value is calculated as follows. An error is read from the error buffer 201 that holds the signal obtained from the error value calculation means 103 together with the error position information, and a predetermined error matrix (in the embodiment shown in FIG. The product of the error matrix and the corresponding value is calculated for all error matrices, and the sum is used as the correction value. In addition, the position shown by x in FIG. 3B means the target pixel. For example, if the quantization error value of a pixel immediately above one line of the target pixel is 32, the value corresponding to that pixel in the error matrix is 4/32; The conversion error is 4 which is the product of both. In this way, the quantization error in a total of 17 pixels, that is, 7 pixels on 2 lines, 7 pixels on 1 line, and 3 pixels on the same line, is read from the error buffer 201 for one pixel of interest, and is multiplied by the error matrix. By performing a sum operation, a sum of errors relating to the pixel of interest is calculated and used as a correction value. Note that the error matrix in FIG. 3B is designed to be 1 when all elements are added. This is because the generated quantization error is used in surrounding pixels without excess or deficiency.

  The error matrix used here is shown in FIG. 3B, but an error matrix as shown in FIG. 3C may be used. The error matrix of FIG. 3C is obtained by multiplying each value in the matrix of FIG. For example, when the quantization error is 32 in the pixel immediately above the target pixel, the value corresponding to that pixel in the error matrix is 4, and therefore the quantization error related to the target pixel from the pixel is The product of both is 128. In this manner, the quantization error of 17 pixels in total, that is, 7 pixels on 2 lines, 7 pixels on 1 line, and 3 pixels on the same line, is read from the error buffer 201 for one pixel of interest, and is calculated as an error matrix. To calculate the error sum related to the pixel of interest. Thereafter, the error sum is divided by 32. By adopting such a method, it becomes possible to calculate the error sum related to the pixel of interest by high-speed shift calculation or integer calculation, and the circuit can be configured at low cost.

  The threshold matrix determination unit 104 switches the threshold matrix according to the gradation value of the target pixel. FIG. 4 shows the configuration of the threshold matrix determining means 104. Each of the threshold matrices 1 to 5 is stored in the storage unit, and the selector 301 selects an optimum threshold matrix according to the output from the image input unit 101, that is, the gradation value of the pixel. This switching is performed according to the input gradation value as shown in FIG. Each threshold value matrix is set as shown in FIG. The effect by switching each threshold value matrix demonstrated here is explained in full detail behind.

  The threshold value determining means 105 determines the threshold value as follows. First, the position on the threshold value matrix corresponding to the target pixel is calculated. That is, when the threshold value matrix is repeatedly tiled on the output image, it is determined according to which position in the threshold value matrix the target pixel corresponds to. Here, in this embodiment, the size of the threshold value matrix is assumed to be constant regardless of the input data condition. A specific position calculation method on the threshold matrix is as follows. When the threshold matrix size is horizontal w pixels and vertical h pixels, the horizontal axis (X mod w) and vertical axis (Y mod) are used as the threshold matrix coordinates for the target pixel of horizontal X and vertical Y in the output image coordinates. The position threshold shown in h) is used. Here, mod is a remainder operator, and (X mod w) indicates a remainder when X is divided by w. For example, since the threshold matrix in this embodiment is w = h = 4, for the pixel of (X, Y) = (9, 6) in the output image coordinates, (x, y) = The threshold value of the position (1, 2) is used. In this case, the threshold value (A), threshold value (B), and threshold value (C) for determining whether or not to output dots corresponding to the output gradation values 85, 170, and 255 are the threshold values in the threshold value matrix 5, for example. From the values at the corresponding positions in the matrix (A), threshold matrix (B), and threshold matrix (C), 176, 192, and 207 are obtained. Thus, by always keeping the size of the threshold matrix constant, the calculation necessary for calculating the position on the threshold matrix is simplified and the cost is reduced.

  The output tone value determining means 106 compares the threshold value determined by the threshold value determining means 105 with the corrected input value output from the corrected input value calculating means 102 according to the processing flow shown in FIG. Determine the gradation value. Hereinafter, a method in which the output gradation value determining unit 106 determines the output gradation value will be described with reference to FIG.

  The magnitude relationship between the corrected input value output from the corrected input value calculating means 102 and the threshold value C determined by the threshold value determining means 105 is compared (step S1). If the corrected input value is larger than the threshold value (C), 255 is set as the output gradation value (step S2). When the corrected input value is not larger than the threshold value (C), the magnitude relationship between the corrected input value and the threshold value B determined by the threshold value determining means 105 is compared (step S3). If the corrected input value is larger than the threshold value (B), 170 is set as the output gradation value. (Step S4). If the corrected input value is not larger than the threshold value (B), the magnitude relationship between the corrected input value and the threshold value A determined by the threshold value determining means 105 is compared (step S5). If the corrected input value is larger than the threshold value (A), 85 is set as the output gradation value (step S6). If the corrected input value is not larger than the threshold value (A), 0 is set as the output gradation value (step S7).

  Each threshold value matrix will be described in detail. The threshold value matrix 5 in FIG. 5B is a dot with less than an isolated gradation value of 170, with halftone dots growing in the order shown in FIG. 3D when the error diffusion matrix in FIG. 3B is used. Is designed so that the average quantization error is close to zero. An isolated dot means a dot in which dots other than the gradation value 0 are not adjacent to the upper, lower, left, and right sides of the dot. However, when the gradation value of the pixel is 0, it is not called an isolated dot. In general, it is difficult for an electrophotographic printer to stably reproduce the density of isolated low density dots. Therefore, in this embodiment, stable density reproduction is aimed at by suppressing the appearance of isolated dots having an isolated gradation value of less than 170.

  Next, the average quantization error will be described in detail. The average quantization error is a value obtained by averaging the quantization error occurring at each pixel position, that is, the value obtained by subtracting the output gradation value from the corrected input value. In the present embodiment, the threshold value matrix is designed so that the average quantization error is substantially constant. The reason for making it substantially constant will be described with reference to FIG. The solid line in FIG. 7 indicates the average quantization error. As shown in FIG. 7, it is assumed that there are two regions, region A and region B, in which the average quantization error is greatly different even though the gradation value hardly changes. Since the average quantization error is small in the region A, even if the process moves to the region B, the predetermined error is not easily reached. As a result, white spots occur in the region B. Furthermore, since the area | region where a dot becomes dense also generate | occur | produces, a granularity deteriorates. On the other hand, when the process shifts from a place where the average quantization error is high to a place where the average quantization error is low, dots are intensively generated in the area immediately after the shift. As described above, if there is an extreme difference in the average quantization error for each region, dot density occurs and the graininess deteriorates. Therefore, it is desirable that the average quantization error is substantially constant.

  Further, in this embodiment, the average quantization error is designed to be approximately zero. This is because when the input gradation value of a region is 0 or 255, the average quantization error in that region becomes 0. Therefore, the average quantization error is almost equal to other input gradation values. This is because 0 is desirable. In order to design such a threshold matrix, the following method (Japanese Patent Application No. 2004-295189) was used.

  First, a corrected input value at each pixel position in a steady state of an output dot pattern to be output to the image forming apparatus is calculated. Here, the steady state refers to a state in which the same pattern is stably generated repeatedly as a result of halftone processing of a solid image having a constant input gradation value. From the corrected input value at each pixel position in the steady state, the range of the threshold value for the given output dot pattern is obtained. This procedure is performed for various output dot patterns, and the threshold range obtained for each pattern is narrowed down.

  FIG. 8 is a flowchart for explaining a threshold matrix design procedure. First, a plurality or a plurality of output dot patterns to be output to the image forming apparatus are prepared. Here, it is assumed that N output dot patterns are prepared. First, one output dot pattern is selected from the N output dot patterns (step S11), and the corrected input value at each pixel position when the selected output dot pattern is in a steady state is obtained. Calculate (step S12). From the calculated corrected input value, a threshold range for obtaining the given dot pattern is obtained (step S13). It is determined whether or not steps S11 to 13 have been completed for all N output dot patterns (step S14). If not completed (No in step S14), the process returns to step S11, and if completed (step S14). Yes) ends. Here, based on the threshold range obtained for each output dot pattern, the final threshold range is narrowed down to determine the threshold. These steps will be described in more detail.

  FIG. 9 shows an example of a set of output dot patterns. A method of calculating the corrected input value at each pixel position in the steady state will be described using the output dot patterns 1101 to 1125 shown in FIG. In FIG. 9, the x direction of each matrix is the main scanning direction, and the y direction is the sub scanning direction. That is, image processing is performed by scanning the laser from the left to the right of each matrix.

  Each output dot pattern in FIG. 9 represents a quaternary output dot pattern in which the output gradation value of each pixel is one of 0, 85, 170, and 255. For example, in the eleventh output dot pattern 1111 in FIG. 9, the average value of the output image data values is (0 × 8 + 85 × 2 + 255 × 6) ÷ 16 = 106.25, so that the gradation value is 106. When 25 large solid images are input, repetition of the output dot pattern indicated by the output dot pattern 1111 appears in a steady state. Here, the calculation of the average value is 8 in the matrix, 2 in the 85, 6 in the 255, and the number obtained by adding the respective values to the number of matrices. Divided by is the average value.

  FIG. 10A is a diagram for describing a method of representing pixel positions in an output dot pattern. The pixel position (x, y) = (0, 0) in the output dot pattern is represented as p0, the pixel position (x, y) = (1, 0) is represented as p1, and the like. Here, when the output gradation value of p0 is b0, the error value generated at p0 is e0, the corrected input value at p0 is f0, and the input gradation value is d, (Equation 1) holds.

FIG. 10B is a diagram in which output dot patterns are arranged. The meaning of Equation 1 is, for example, that the corrected input value f0 at p0 is obtained by adding the input tone value and the value weighted to the error value from the shaded pixel position in FIG. From that
f0 = d + 1/32 × (0, e3 + 1, e0 + 1, e1 + 2, e2 + 1, e3 + 1, e0 + 0, e1 + 1, e7 + 1, e4 + 2, e5 + 4, e6 + 2, e7 + 1, e4 + 1, e5 + 2, e1 + 4, e2 + 8, e3) = d + 1/32 × (2 E0 + 3, e1 + 6, e2 + 9, e3 + 2, e4 + 3, e5 + 4, e6 + 3, e7). This can be confirmed in the first row of (Equation 1).

  Further, since the error value e0 at p0 is the difference between the corrected input value and the output gradation value, the equation fn = en + bn (where 0 ≦ n ≦ 7) is obtained from fn−bn = en (equation 2). Thus, substituting into Equation 1 yields (Equation 3).

（式３）からは、誤差値ｅｎや修正入力値ｆｎにはそれぞれ無数の解が存在するので、一意に求めることができない。一意の解を得るためには、条件式を加える必要があるが、グラデーション画像に対する階調段差抑制を図るため、誤差の平均値を条件として与える。

From (Equation 3), there are innumerable solutions for the error value en and the corrected input value fn, respectively. In order to obtain a unique solution, it is necessary to add a conditional expression, but in order to suppress a gradation step for a gradation image, an average error value is given as a condition.

Here, the average value of errors is given as zero. This is equivalent to replacing Equation 4 where the sum of error values is 0 with one of the rows in Equation 3.
0 = e0 + e1 + e2 + e3 + e4 + e5 + e6 + e7 (formula 4)
An example in which the top line is replaced is shown in Equation 5.

ここで、ｄ、ｂｎは定常状態の出力ドットパターンを与えることから、いずれも既知の定数となっているので、誤差値ｅｎが求まる。上記の場合は、最上位の行を置き換えたが、この理由は上述した無数の解が存在することから一意に求めるには、さらなる条件を加える必要があるからである。式３の中にある８つの行のうち、１つの行は意味をなしていない式であり、つまり、他の７行の式から求まる式であるからである。

Here, since d and bn give a steady-state output dot pattern, since both are known constants, the error value en is obtained. In the above case, the top row has been replaced because the above-mentioned innumerable solutions exist, and it is necessary to add further conditions to obtain uniquely. This is because one of the eight lines in the expression 3 is an expression that does not make sense, that is, an expression obtained from the expressions in the other seven lines.

  FIG. 11 is a diagram for explaining a set of corrected input values corresponding to the set of output dot patterns in FIG. The corrected input values 1701 to 1725 are calculated from Equation 1 or Equation 2 based on the error value en. For example, when the output dot pattern 1111 in FIG. 9 is given, the corrected input value indicated by 1711 in FIG. 11 is obtained.

  Here, in order to obtain the output dot pattern given in FIG. 9, for example, at p0, the corrected input values indicated by output gradation values 0 and 1711 are the correction input values indicated by 1701 to 1710 in FIG. For the corrected input values indicated by the output gradation values 85 and 1712, the output gradation value 170 must be a threshold value for generating the output gradation value 255 for the corrected input values indicated by 1713 to 1725. There is. Therefore, the threshold A for determining whether or not to output the output gradation value 85 at p0 is not less than the maximum value 121.2 of the corrected input value from 1701 to 1710 and the minimum value 136.5 of the corrected input value from 1711 to 1725. Need to be smaller. Therefore, a value 129 obtained by rounding off the decimal point in the median value of the numerical values of 121.2 or more and less than 136.5 is set as the threshold A at p0.

  Similarly, the threshold value B for determining whether or not to output the output gradation value 170 is not less than the maximum value 136.5 of the corrected input value 1701 to 1711, and from the minimum value 151.8 of the corrected input value 1712 to 1725. Since it needs to be small, it is set to 144. Similarly, the threshold value C for determining whether or not to output the output gradation value 255 is greater than or equal to the maximum value 151.8 of the corrected input value from 1701 to 1712 and from the minimum value 167.0 of the corrected input value from 1713 to 1725. Since it needs to be small, it is set to 159, which is the median value of 151.8 and 167.0, rounded down after the decimal point. In this way, the thresholds A to C are obtained.

  As is clear from the above method, the optimum threshold matrix obtained here is uniquely determined. Here, in the threshold matrix 5 of FIG. 5B, the threshold matrix (A), the threshold matrix (B), and the threshold matrix (C) are dots corresponding to the output gradation value 85, dots corresponding to 170, This is a matrix having a threshold value for determining whether or not to output a dot corresponding to 255 as an element, and a threshold value matrix for quaternary output expressing a halftone dot screen of about 212 lines and 45 degrees at an output resolution of 600 dpi. That is, the output gradation value is determined from among four types of output gradation values, 0, 85, 170, and 255. When the threshold value matrix 5 of FIG. 5B is used, for an input image having a low gradation value, for example, an input image composed of only the gradation value 18, approximately (x, y) = (1, The gradation value 0 or 170 is output only at the pixel positions corresponding to 1) and (3, 3), and only the gradation value 0 is output at the other pixel positions.

  Here, the degree of freedom will be described as an index indicating whether the dots are likely to come out in a dispersed manner or at a specific position. The degree of freedom is defined as the ratio of the number of pixels whose output gradation value is selected from a plurality of values to the number of pixels in the threshold value matrix (16 in this embodiment). In the case of the threshold matrix 5, the output gradation value is selected from a plurality of values in approximately 2 pixels per 16 pixels, and is fixed at a specific value, here 0 in the other 14 pixels. The degree is expressed as 2/16.

  Next, the threshold matrix 4 will be described. Similar to the threshold matrix 5, the threshold matrix (A), the threshold matrix (B), and the threshold matrix (C) in the threshold matrix 4 in FIG. 5B are dots corresponding to the output gradation value 85 and 170, respectively. This is a threshold value for determining whether or not to output dots corresponding to 255 and dots corresponding to 255. By using the threshold matrix 4 for an input image having a relatively low gradation value, the output image is generally expressed by one of output gradation values 0 and 170. Also, in the pixel positions corresponding to (x, y) = (1, 0), (0, 1), (1, 1), (3, 2), (2, 3), (3, 3). The output gradation value is 0 or 170, and only 0 appears at the other pixel positions. That is, when an image having a low gradation value is input, when a threshold value based on the threshold value matrix 4 is used, a dot having a gradation value 170 appears at a pixel position where the dot having the gradation value 170 appears and the surrounding pixel positions. It will be. Since the output gradation value is generally selected from a plurality of values with 6 pixels per 16 pixels, the degree of freedom of the threshold value matrix 4 can be expressed as 6/16.

  As described above, since the threshold matrix 5 has a very low degree of freedom, it is easy to make a predetermined screen. On the other hand, since the threshold matrices 1 to 3 have a very high degree of freedom (degree of freedom 16/16), the dot dispersibility is high. By interposing the threshold value matrix 4 having a degree of freedom of 6/16, it is possible to smoothly connect a density range where a screen can be easily created and a density range where dot dispersibility is high, for the purpose of the present invention. As described above, the uncomfortable feeling caused by the boundary between the two can be reduced, and the image quality is improved.

  Here, comparing the case where the threshold matrix 4 is used with the case where the threshold matrix 5 is used, the case where the threshold matrix 4 is used increases the degree of freedom of the gradation value distribution of the output image, and the gradation It is easy to disperse dots having a value of 170. When the threshold matrix 4 is used, there is a possibility that the output gradation value is 255. However, in order to do so, the corrected input value calculated by the corrected input value calculation means 102, that is, the sum of the diffused error and the pixel value needs to exceed 255, and the probability is low. When the input tone value of the target pixel is low, the probability is low. Further, when the threshold value based on the threshold value matrix 4 is used, as is apparent from the processing flow of FIG.

  As clarified in the processing flow of FIG. 6, the correction input value and the threshold value matrix (C) where the gradation value is 255, the threshold value matrix (B) where the gradation value becomes 170, and the threshold value matrix (A) where the gradation value becomes 85 are set. The threshold values in the threshold matrix are compared. Since the values in the threshold matrices in (A) and (B) are exactly the same, the gradation value 85 is not output in the pixel where the gradation value 170 is not output. . That is, an isolated dot having a gradation value less than 170 does not appear.

  The threshold matrix 4 has a distortion around the input gradation value of 22 when it is desired to suppress distortion at the switching portion with the region using the threshold based on the threshold matrix 5, and at the switching portion with the region using the threshold based on the threshold matrix 3. When it is desired to suppress, it is designed that the average quantization error is close to 0 with respect to an input image composed only of gradation values around the input gradation value of 18 so that the average quantization error does not change greatly between processes. Therefore, it is preferable. That is, when the average quantization error changes greatly, image quality degradation, that is, distortion occurs at the boundary.

  The reason is as follows. Dither threshold error diffusion processing and error diffusion processing are switched according to image characteristics to obtain a halftone processing result that takes advantage of both methods. By changing the matrix in a manner similar to the matrix, the matrix size of the diffusion error matrix is reduced and the elements of the matrix are reduced. On the other hand, the threshold value is set to a constant value in the vicinity of the image suitable for the error diffusion method. Thus, an image processing apparatus has been proposed in which the matrix size of the diffusion error matrix is increased to increase the matrix elements (see, for example, Patent Document 2). In this image processing apparatus, when the image data is suitable for the dither method, the threshold value is changed in a manner similar to the dither matrix, the matrix size of the diffusion error matrix is reduced, and the elements of the matrix are reduced. However, when halftone processing is performed at the boundary between the area suitable for the dither method and the area suitable for the error diffusion method, the processing method is different at the boundary. Image quality degradation such as omission occurs. This deterioration in image quality occurs because the average quantization error is not constant, as will be described later.

  FIG. 12A shows a document image input to inspect an image state by a conventional dither threshold error diffusion processing method. 12 (b) to 12 (d) are diagrams showing character area images processed by the conventional dither threshold error diffusion processing method by reading the original image shown in FIG. 12 (a). Here, the threshold value is changed in a manner similar to a dither matrix inside the character region in the image, and binarization processing is performed with the threshold value kept constant at the boundary of the character region.

  FIG. 12A shows an input document image, the background has a gradation value of 0, and the character gradation values are 63 for image 1, 127 for image 2, and 192 for image 3. The gradation value takes an integer value from 0 to 255, and the larger the value, the higher the density. The dither matrix used is designed so that the average quantization error is zero.

  FIGS. 12B, 12C, and 12D show images obtained as a result of processing the images with the threshold values 81, 127, and 173 at the boundary of the character area, respectively. As will be described later, by setting the threshold 81 when the input gradation value is 63, setting the threshold 127 when the input gradation value is 127, and setting the threshold 173 when the input gradation value is 192, the average quantization error is substantially reduced. 0.

  Here, in the image 1 in FIG. 12B, the image 2 in FIG. 12C, and the image 3 in FIG. 12D in which the average quantization error is substantially constant 0 in the entire area of the image, The image has no white spots or blurs and almost no noise.

  Compared with these, when the average quantization error is not substantially zero at the edge of the character, the following image quality degradation occurs. That is, as shown in images 2 and 3 in FIG. 12B and image 3 in FIG. 12C, white spots occur inside the characters. In addition, as in the image 1 in FIG. 12C and the images 1 and 2 in FIG. 12D, a phenomenon occurs in which the characters are blurred due to the dots being hit in the background portion close to the characters.

  Next, the threshold matrices 1 to 3 will be described in detail. As is apparent from FIG. 5B, the threshold values 1 to 3 have the same threshold value (A) and threshold value (B). Therefore, the output gradation value never becomes 85, and the output gradation value 170 or 255 is selected, and it is impossible that a small dot is isolated and output. Further, the threshold value (C) is uniformly 255, and the threshold value matrices 1 to 3 are selected when the pixel of interest has a relatively low density. Therefore, the corrected input value, that is, the diffused error and the pixel value Since it is unlikely that the sum exceeds 255 over 255, it is unlikely that 255 will actually be output. Further, since each threshold matrix is uniform, it can be said that it is an error diffusion process with an output gradation value of 170 substantially, and its output dot arrangement is a non-periodic dispersion type. When a substantially uniform threshold value matrix is used in this way, the amount of memory held as the threshold value matrix can be greatly reduced, so that the purpose of generating a distributed dot arrangement is achieved, but the amount of memory required for it is minimized. This will save costs.

  Further, the threshold values (A) and (B) of the threshold matrices 1 to 3 are determined based on FIG. As will be described later, since FIG. 13 presents a threshold value matrix in which the average quantization error is approximately 0, the use of the threshold value matrix with reference to FIG. 13 does not cause the above-described defect related to the average quantization error. In the embodiment, three types of threshold matrixes are prepared based on FIG. 13. However, if a larger number of threshold matrices are prepared, the average quantization error becomes closer to 0 and a favorable result is obtained. Furthermore, the structure which adds blue noise etc. with respect to the threshold value based on the threshold value matrices 1-3 may be sufficient.

  By the way, in the process using the threshold value based on the threshold value matrix 5, since a halftone screen having a desired number of lines can be formed for an image having a high input tone value, the graininess is excellent, but the input tone value is low. In an image, the number of dots required to form a desired number of lines is not sufficient, and the graininess deteriorates. This is because the latter is an image containing a lot of low-frequency components and appears to be rough to the human eye.

A halftone dot screen having an isolated minimum dot having a gradation value of 170 or more and an output resolution of 600 dpi of about 212 lines and 45 degrees has a gradation value 170 of 2 pixels per 16 pixels, as shown in FIG. 14 pixels are filled with dots having a gradation value of 0, and the average gradation value at this time is 21.25 from Equation 6.
{170 × 2 + 0 × 14} ÷ 16 = 21.25 (Formula 6)
That is, this screen cannot be made for an input image having an average gradation value of less than 21.25. For such an input image, it is often the case that using a process in which dots are evenly dispersed results in less roughness and better graininess.

  Therefore, processing is performed using a threshold value based on the threshold matrix 5 for a region with a high input tone value, and dots are dispersed for a region with a low input tone value. By performing processing using the threshold value based on this, it is possible to obtain an image with excellent graininess for various input tone values. Then, for a region having a gradation value between the two, processing is performed using a threshold value based on the threshold value matrix 4 in which dots are dispersed moderately. When switching processes with greatly different dot dispersion, the process switching section is likely to be distorted for gradation images with gentle gradation changes, but use a threshold value for moderate dot dispersion. Thus, this distortion can be reduced.

  Further, in the case of using a threshold value based on the threshold value matrix 4, when a threshold value based on the threshold value matrix 5 is used, a dot having a gradation value 170 is generated only at the pixel position where the dot having the gradation value 170 appears and the surrounding pixel positions. As a result, distortion at the process switching unit between the two can be reduced. Which threshold value to use for which input gradation value is determined by outputting an image at each threshold value so that the roughness is small and the distortion of the processing switching unit is not noticeable. At the same time, the threshold matrix 4 is determined so that the distortion of the process switching unit is not noticeable.

  Next, a method for designing the threshold matrices 1 to 3 will be described with reference to FIG. First, in order to obtain a threshold value for the input tone value d, a solid image having a sufficiently large input tone value d is prepared as an input image. Here, an image of 1000 dots square indicated by 901 in FIG. Next, a temporary threshold value is given. The threshold A and threshold B are assumed to be temporary values t, and the value of the threshold C is fixed at 255. Next, halftone processing using a temporary threshold is performed on the prepared solid image, and an average quantization error occurring in a predetermined range is obtained. Here, an average quantization error in a 255 dot square area indicated by 902 in FIG. 15 is obtained. By changing the temporary value t from 0 to 255 in increments of 1 and repeating the above operation, a threshold value with which the average quantization error is closest to 0 with respect to the input gradation value d is obtained. Further, by repeating the above operation while changing the input gradation value d, the relationship of the threshold values shown in FIG. 13 where the average quantization error corresponding to the input gradation value is closest to 0 is obtained. The curve indicating the relationship between the input gradation value and the threshold value in FIG. 13 is corrected so as to satisfy the relationship of Equation 7 gently.

t (d + 1) ≧ t (d) (Formula 7)
However, d is an input gradation value, and t (d) is a threshold for the input gradation value d. Further, instead of the threshold matrix 4 shown in FIG. 5B, a configuration using the threshold matrix 4 ′ in FIG. 16 in which some values of the threshold matrix are changed may be used. Further, a configuration using a threshold value matrix 5 ′ shown in FIG. 16 instead of the threshold value matrix 5 shown in FIG. That is, (x, y) = (1, 0), (0, 1), (1, 1), (3, 2), (2, 3), (3, 3) in the threshold matrix are set to the same value. It is not necessary to set the values independently, and the probability that a dot having a gradation value of 170 will be generated at the corresponding pixel position can be controlled. For example, the threshold value can be set so that the probability of dot generation at each pixel position is the same. Further, when the threshold matrix 5 is used, a dot having the gradation value 170 is most likely to appear (x, y) = (1, 1), (3, 3). Distortion at the boundary where the threshold matrix 4 and the threshold matrix 5 are switched can be suppressed.

  In this embodiment, the optimum threshold value matrix is used for the medium density region in addition to the optimum threshold value matrix for the low density region and the high density region. However, depending on the output printer, the low density region and the high density region are originally used. In such a case, if the configuration does not use an optimum threshold value matrix for the medium density region, the required memory amount can be reduced and the cost can be reduced.

1 shows a general laser printer to which the present invention is applied. The structure of the halftone process of this invention is shown. The structure of a correction input value calculation means is shown. The structure of a threshold value matrix determination means is shown. The threshold value matrix switched according to an input gradation value is shown. It is a process flowchart which determines an output gradation value. It is a figure explaining an average quantization error. It is a flowchart explaining the design procedure of a threshold value matrix. An example of a set of output dot patterns is shown. It is a figure explaining the notation method of the pixel position in an output dot pattern. It is a figure explaining the group of the correction input value corresponding to the group of the output dot pattern of FIG. It is a figure explaining the reason which makes an average quantization error zero. The correspondence relationship between the input gradation value and the threshold value, in which the average quantization error is close to 0, is shown. An example of a halftone screen is shown. It is a figure explaining the design method of the threshold value matrices 1-3. The other example of a threshold value matrix is shown. It is a figure explaining the subject of a prior art.

Explanation of symbols

101 Image Input Unit 102 Modified Input Value Calculation Unit 103 Error Value Calculation Unit 104 Threshold Matrix Determination Unit 105 Threshold Determination Unit 106 Output Tone Value Determination Unit

Claims (5)

A corrected input value calculating means for calculating a corrected input value by adding a predetermined weight to an error value around the target pixel position of the image information to be input and calculating the corrected input value, or an input value of the target pixel from a plurality of threshold matrixes, or Threshold value determining means for determining a threshold matrix to be applied based on an input value around the target pixel; threshold determining means for determining a threshold value to be applied based on the determined threshold matrix; and the target pixel position; An output tone value determining unit that determines an output tone value based on the determined threshold for the corrected input value of the target pixel position calculated by the corrected input value calculating unit, and the output tone value determining unit determines An image processing unit comprising: an error value calculating unit that calculates a difference between the output gradation value and the corrected input value as the error value, and transmits the calculated error value to the corrected input value calculating unit. An apparatus, wherein the plurality of threshold matrix state, and are not the average quantization error becomes substantially constant value in the input tone values each threshold matrix is used, the threshold value matrix determination unit, in a low concentration range Select a threshold matrix to be a distributed dot arrangement, select a threshold matrix to be a periodic dot arrangement in the high density area, and between the low density area and the high density area, an intermediate dot arrangement between the two An image processing apparatus characterized by selecting a threshold matrix . A corrected input value calculation step of calculating a corrected input value by adding a predetermined weight to an error value around the target pixel position of the image information to be input and calculating the corrected input value, or an input value of the target pixel from a plurality of threshold matrices, or A threshold value matrix determining step for determining a threshold value matrix to be applied based on an input value around the target pixel; a threshold value determining step for determining a threshold value to be applied based on the determined threshold value matrix and the target pixel position; An output tone value determining step for determining an output tone value based on the determined threshold value for the corrected input value of the target pixel position calculated by the corrected input value calculating step, and the output tone value determining step are determined An error value calculating step of calculating a difference between the output gradation value and the corrected input value as the error value, and transmitting the calculated error value to the corrected input value calculating step. A method, wherein the plurality of threshold matrix state, and are not the average quantization error in the input tone value each threshold matrix used is almost a constant value, the threshold value matrix determining step, in the low density region Select a threshold matrix to be a distributed dot arrangement, select a threshold matrix to be a periodic dot arrangement in the high density area, and between the low density area and the high density area, an intermediate dot arrangement between the two An image processing method characterized by selecting a threshold matrix . A program for causing a computer to implement the image processing method according to claim 2 . A computer-readable recording medium recording a program for causing a computer to implement the image processing method according to claim 2 . An image forming apparatus comprising the image processing apparatus according to claim 1 .

Images