JP3374551B2 - Image processing device - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus,
In particular, the present invention relates to an image processing apparatus using an error diffusion binarization method. 2. Description of the Related Art Conventionally, there is an error diffusion system as a binarization system for achieving both gradation and resolution of an image processing apparatus. In the error diffusion method, the multi-valued image subjected to the error correction is binarized by a threshold value, and the binarization error generated at that time is propagated to the periphery to preserve the area gradation. This method is R.
Proposed by Floyd et al. FIG. 6 is a block diagram showing a configuration of an image processing circuit 31 for realizing a general error diffusion method employed in a conventional image processing apparatus. Referring to FIG. 6, a conventional image processing circuit inputs an image data represented by 8 bits and performs an error correction, and an error correction unit 22 that inputs the error-corrected input image data to a predetermined threshold Th. , A binarization error calculator 24 that calculates an error between the binarization output data and the error-corrected input image data, and a correction for diffusing the binarization error. An error correction is performed on the input image data by using the correction data prepared by the binarization error diffusion unit 25. A binarization error calculator 24 receives the binarized output image data and two reference values, low and high, and
A selector 241 that outputs one of the reference values based on the value output image data, and a binarization error calculator 242 that calculates a binarization error from the data output from the selector 241 and the error-corrected input image data. And The binarization error diffusion unit 25 includes a first-in first-out error storage line memory 251 for storing two lines of data output from the binarization error calculation unit 24,
An error weighting filter 2 connected to the error storage line memory 251 and performing greater weighting closer to the pixel of interest
52. Next, the operation will be described. Let the density of the input multi-valued image be f (x, y) (0 ≦ f (x, y) ≦ 1)
If the output image density is g (x, y) (= 0 or 1),
The binarization error Exy is represented by the following equation (1). Exy = f (x, y) −g (x, y) (1) The error diffusion method is to average the Exy and reduce the Exy. Correction is performed using the weighted average value. If the weighted average value Eave _xy, the weighted average value Eave _xy is represented by the following formula (2). [0007] In general, the weighting coefficient _{mk, n} becomes larger as it approaches the pixel of interest, and the matrix size k in the main scanning direction is 5
Pixels and the size n in the sub-scanning direction are often 2 to 3 lines. A specific example of such an error weighting filter 252 is as shown in FIG. Referring to FIG. 6, the conventional image processing apparatus forms a feedback loop to diffuse a binarization error. Because of the feedback loop, it is necessary to pay particular attention to the conditions at the leading or trailing edge of the image (initialization and effective image area setting). Further, since a series of processes cannot be pipelined, it can be said that it is basically difficult to increase the speed. Here, for example, in the case of a white low-density thin line such as a hand-drawn pencil manuscript, the amount of occurrence of the binarization error is small and the width of the line drawing is narrow, so that the error cannot be propagated in time.
The image is interrupted, and at the same time, a ghost error occurs. If the ghost error is accumulated and exceeds the binarization threshold value outside the line drawing area, line drawing distortion occurs. The discontinuity and distortion of the low-density thin line are the biggest defects caused by the principle of the error diffusion method. A directional problem in a boundary region where the density rapidly changes from a low density to a high density (or vice versa) also occurs for the same reason. If the underlayer is covered to some extent, these phenomena are unlikely to occur. In the solid area, an irregular texture is generated (which is closely related to the shape of the error weighting filter). In particular, in a low-concentration part, the texture is conspicuous, and if you are not familiar with it, you may feel strange. Further, in the intermediate density portion, the high frequency texture component is easily crushed at the time of printing, which may result in black spots. FIGS. 7 to 9 show examples of images processed by the conventional image processing circuit shown in FIG. Referring to FIG. 7, in the low-density thin line, breaks and rattling occur (arrows in FIG. 7, threshold value: 128/255, which is binarized. Improvements are possible). Referring to FIG. 8, irregular texture is conspicuous in the low density portion. There are also black spots. Referring to FIG. 9, the direction between the black image and the white image has remarkable directionality such as black cast or white drop. Here, the texture means a stripe pattern. FIG. 10 is a block diagram showing the configuration of an image processing circuit for realizing a binarized average density error diffusion method which solves the problems of the conventional general error diffusion method. Referring to FIG. 10, image processing circuit 3 for implementing the binarized density error diffusion method
6 is characterized in that it has a second loop for diffusing the error between the average density of the output binary image and the original image in addition to the first feedback loop for diffusing the error shown in FIG. That is, referring to FIG. 10, in addition to image processing circuit 31, image processing circuit 32 outputs an error between the average density of the output binary image and the original image with respect to the output data from binarization error calculating section 24. And a subtraction unit 2 for subtracting the binary averaged error data from the binary averaged error diffusion unit 26 from the input image data.
The error correction unit 22 corrects the input image data in consideration of the subtracted image data. The binarized average error diffusion unit 26 includes an output binary image storage line memory 261 for storing two lines of binary image data, and an output 2
And a binarized average density weighting filter 262 connected to the value image storage line memory 261 and performing predetermined weighting on the output binary image. If the binary average value is Bave _xy in the binary average density error diffusion method, the binary average value Bave _xy is expressed by the following equation (3). [Equation 2] The binary average Bave _xy and the original image f
Assuming that the difference between (x, y) is EBave _xy , 2
The quantified average error is represented. This is input multi-valued image density f
Correct (x, y). EBave _xy = f (x, y) −Bave _{x, y} (4) An error occurs in the second loop because the difference between the original image and the average binary output density is as shown below. When the following phenomenon occurs in the first loop, the second
Works in a direction to correct it. (I) The binary output is interrupted / distorted by a low-density thin line or the like. (Ii) Directivity has occurred. (Iii) An extremely low frequency texture is generated. In other words, when the principle of error diffusion occurs in the first error diffusion loop, the processing error is absorbed / diffused by the second loop, and high image quality is achieved. FIGS. 11 to 13 show processing samples according to this method. Referring to FIG. 11, in the low-density thin line, the interruption / distortion is largely improved by the correction effect of the second loop described above. Referring to FIG. 12, there is little improvement effect on black spots. Referring to FIG. 13, a considerable improvement effect can be seen in the directionality. Therefore, the present method has almost the same or better image quality than the general error diffusion method in most image attributes. However, since the number of feedback loops increases in terms of hardware, the operation speed is reduced. Next, a dither error diffusion method will be described as another method for improving the conventional general error diffusion method. FIG. 14 is a block diagram showing a configuration of an image processing circuit for realizing the dither error diffusion method. Referring to FIG. 14, an image processing circuit 33 that implements a dither error diffusion method uses a 4 × 4 dither matrix for 8-bit input image data with respect to a conventional image processing circuit 31 and uses a 4-bit dither matrix. Image data is output, and error correction is performed on the output image data. Referring to FIG. 14, multi-valued dither section 21
The adder 211 that inputs the upper 4 bits of the bit input image data and “1” and outputs the sum thereof, and counts the horizontal / vertical synchronization signals in the raster scan by a counter (not shown), and outputs 4 by a decoder (not shown). A dither matrix 212 that outputs a 4-bit threshold value of a × 4 dither matrix is compared with a 4-bit threshold value Txy output from the dither matrix 212 and lower 4 bits of 8-bit input image data, and 1 A comparator 213 that outputs bit data, upper 4 bits of 8-bit input image data, and addition data from the adder 211 are input, and any one of the data is selectively selected according to a signal from the comparator 213. And a selector 214 for outputting the result. In the dither error diffusion method, the operation speed can be greatly improved and the cost can be reduced by hardening the error diffusion method. By the way, in order to realize the conventional 256 gradations for 8-bit input image data, it is necessary to set the basic operation data width of the feedback section to 8 bits. If it is 8 bits, the coefficient 16
Requires an addition operation of up to 12 bits. However, if this method is used, the basic operation data width of 4 bits is enough to realize 256 gradations, and the maximum addition operation described above requires only 8 bits. As a result, a significant improvement in operation speed can be realized. Also, the error storage memory is half that of the conventional one. Further, in terms of image quality, generation of irregular textures in computer graphic images and the like is suppressed by dither components. The processing samples in this case are shown in FIGS.
Shown in It can be seen that the occurrence of irregular texture is suppressed as compared with FIGS. FIG. 18 is a block diagram showing the structure of an image processing circuit 34 in which the binarized average density error diffusion method shown in FIG. 6 and the dither error diffusion method shown in FIG. 14 are simply combined. Referring to FIG. 18, since the configuration of image processing circuit 34 in which the binarized average error diffusion method and the dither error diffusion method are combined is a combination of the configurations described above, the same reference numerals are assigned to the same portions. The description is omitted here. FIGS. 19 to 21 show sample images obtained by an image processing circuit 34 in which the binarized average error diffusion method and the dither error diffusion method are simply combined. is there. Referring to FIG. 19, in the image processing circuit 34 according to this configuration, the reproducibility of a low-density thin line is significantly improved compared to the image processing circuit 33 of FIG. 14 in which the image processing is performed by the dither error diffusion method alone. I do. However, in an image with a small variance such as a computer graphics image, an irregular stripe pattern (texture) peculiar to error diffusion occurs again due to the effect of correcting the binarized average error. Due to this effect, black spots have recurred. By combining the binarized average error diffusion method and the dither error diffusion method, a high-quality and high-speed error diffusion method can be constructed. However, in a simple combination, a phenomenon may occur in which the image advantages of the two are canceled by each other's algorithm. The present invention has been made in order to solve the above-mentioned problems. Even when the binarized average error diffusion method and the dither error diffusion method are combined, a stripe pattern is generated in computer graphics or the like. It is an object of the present invention to provide an image processing apparatus which does not have any problem and can perform high-speed error diffusion. An image processing apparatus according to the present invention comprises an N2 value conversion means for converting input image data represented by an N1 value into image data having an N2 value smaller than the N1 value. , N3 value converting means for converting image data represented by the N2 value into N3 value data smaller than the N2 value, and first error error diffusion data preparing means for preparing data for diffusing an error in the N3 value conversion. A second error diffusion data preparing unit for preparing second error diffusion data for preparing data for diffusing an error between the average density of the output image at the time of the N3 value conversion and the N2 value image around the pixel of interest; Based on the image data, the maximum and minimum values of the local area of the input image data
Greatest local region according to the difference, and means for determining one of dispersion and edge detection amount, the determination result of the determine by means
Means for adjusting so that the larger the difference between the value and the minimum value, the variance or the detected edge amount is, the larger the feedback of the second error diffusion data is, and the first error diffusion data and the adjusted second error diffusion data are used. Means for correcting the N2 value image data by using The input N1 value image data is converted into N2 value image data, and the difference, variance, and edge difference between the maximum value and the minimum value of the local area of the image are converted based on the N2 value image data.
Any one of the detected amounts is determined. Then, the diffusion amount of the second error diffusion means for diffusing the error between the average image density and the image data is controlled based on the determination result. Station of images
Difference, variance and edge detection between the maximum and minimum values of the area
Since the correction amount of the N3- valued average error is controlled according to any one of the amounts, the maximum value and the minimum value of the local region of the image are
N3 binarization processing according to the difference, variance, or edge detection amount is performed. Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing a configuration of an image processing apparatus according to the present invention. Referring to FIG. 1, an image processing apparatus according to the present invention has a multi-valued dither circuit 11 for inputting 8-bit image data and converting it to 4-bit dither image data, and an output from multi-valued dither circuit 11. An error correction unit 12 for error-diffusion of the obtained 4-bit dither image data with error data accompanying binarization, which will be described later, and an error correction unit 12 for error correction with a predetermined threshold value Th A binarizing unit 13 for binarizing the obtained dither image data, and a binarizing error for calculating a difference between the error-corrected dither image data input to the binarizing unit 13 and the binarized output image data A calculation unit 14, a binarization error diffusion unit 15 for performing error diffusion on the calculated binarization error, and an error between the binarized average density of the binarized output image data and the original image data Binarized average error diffusion section 1 for outputting A binarized average error diffusion unit 16, an attribute determination unit 17 that determines the attribute of the image based on the 4-bit 4 × 4 dither matrix image data output from the multi-valued dither circuit 11, And an adjusting unit 18 for adjusting the binarized average density error according to the determination result from the unit 17. Next, the operation will be described. The input image data f1 is converted into dither image data f2 by the multi-value dither circuit 11, and error-corrected by the error correction unit 12 to become error correction data f3. The error correction data f3 is output by the binarizing unit 13 as a binary output image g1. The binary output image data g1 is converted into one of the reference values according to the value by the binarization error calculation unit 141, and the converted binarization error data g2 is input to the binarization error diffusion unit 15, It is output to the error correction unit 12 as binary error diffusion data h1. The converted binarized error data g2 is also input to the binarized average error diffusion unit 16. Binary average error diffusion unit 16
The binarized average error density data h2 is processed by the adjustment unit 18 described later, and sent to the error correction unit 12 as error correction data h4. The attribute determining section 17 includes a dither storage line memory 171 for storing dither matrix image data for five lines in a first-in first-out format.
The dither storage line memory 171 generates a 5 × 5 local window. The maximum and minimum values in the 5 × 5 local window are detected by the maximum / minimum value detection filter 172, and the maximum / minimum value detection filter 17
The attribute of the image is determined by the gain control circuit 173 connected to the second. Specifically, when the difference between the maximum value and the minimum value is small, the image data is determined to have the non-dispersion attribute, and when the difference is large, it is determined to be the distribution attribute. FIG. 2 is a diagram for explaining a specific operation of the gain control circuit 173. Referring to FIG. 2, the gain is determined according to the difference between the maximum value and the minimum value detected by maximum / minimum value detection filter 172. That is, the gain control circuit 173 includes gain conversion threshold values Th1 and Th2 for dividing the difference between the maximum value and the minimum value into three, and the gain according to the attribute of the image is determined according to the threshold values. You. Specifically, when the difference between the maximum value and the minimum value is smaller than the threshold value Th1, the gain is 0, and the difference is equal to the threshold value Th1.
Is set to 0.5 when the difference is between Th2 and Th2, and set to 1 when the difference is greater than the difference threshold value Th2. Next, referring again to FIG. 1, the adjusting unit 18 outputs the dither image data f2 output from the multi-valued dither circuit 11.
And a subtraction circuit 181 for calculating an error between the average density h2 of the output binary image output from the binarized average error diffusion unit 16 and an error data h3 obtained as a result of the subtraction by the gain control circuit 173. And a multiplication circuit 182 for multiplying the gain according to the attribute of the image. The error diffusion data h4 adjusted in this way is output to the error correction unit 12, and
Error diffusion is performed on the dither image data f2. In the image processing circuit according to the present invention shown in FIG. 1, the maximum value / minimum value of a local region of an image is detected, and the attribute of the image is determined according to the detected value. For example, in a computer graphics image or the like, the difference between the maximum value and the minimum value of the local region is extremely small because the variance of the density is extremely small. The smaller the difference between the maximum and minimum values is, the more 2
Control is performed so that the feedback amount of the quantified average error becomes small. That is, when the variance is extremely small, only the dither error diffusion is performed by canceling the correction of the binarized average density error. As a result, it is possible to prevent the occurrence of a unique stripe pattern in an image with extremely small dispersion of computer graphics or the like. FIGS. 3 to 5 are views showing examples of output images when image processing is performed in the image processing circuit according to the present invention. Since the image area determination is performed with reference to FIGS. 3 to 5, the occurrence of a stripe pattern in a computer graphics image is suppressed, the reproducibility of low-density thin lines is good, and the occurrence of black spots is also suppressed. You can see that In the above embodiment, the attribute of the image is detected by detecting the maximum value and the minimum value of the local area. However, the present invention is not limited to this, and the edge or variance of the image may be detected. At the time of edge detection of an image, when the detection amount is small, the adjustment unit 18 may reduce the error correction amount. When the attribute determination of the image is detected by using the variance, if the variance is small, the adjustment unit may reduce the error correction amount. If the difference is small when the maximum / minimum value is detected in the above embodiment, the adjustment unit 18 may reduce the error correction amount. As described above, according to the present invention, the input N1 value image data is converted into N2 value dither image data, the attribute of the image is determined, and the attribute of the image is determined based on the determination result. Error between the average image density and the dither image data. Since the error diffusion is controlled according to the attributes of the image, image processing is performed which takes advantage of the image processing possessed by both dither error diffusion and binarized average density error diffusion. As a result, while being able to operate at high speed and at low cost, there is little occurrence of discontinuity / typing of low-density thin lines in the conventional error diffusion method, and it is peculiar to image data with little dispersion such as computer graphics images. The generation of the stripe pattern can be suppressed low, and high image quality can be achieved.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram showing a configuration of an image processing circuit according to the present invention. FIG. 2 is a diagram for explaining contents performed in a gain control circuit. FIG. 3 is a diagram showing an output example when an image is processed by the image processing circuit according to the present invention; FIG. 4 is a diagram showing an output example when an image is processed by the image processing circuit according to the present invention; FIG. 5 is a diagram showing an output example when an image is processed by the image processing circuit according to the present invention. FIG. 6 is a block diagram showing a configuration of an image processing circuit for realizing a conventional general error diffusion method. FIG. 7 is a diagram illustrating an example of an output image obtained by a conventional general error diffusion method. FIG. 8 is a diagram showing an example of an output image obtained by a conventional general error diffusion method. FIG. 9 is a diagram showing an example of an output image obtained by a conventional general error diffusion method. FIG. 10 is a block diagram illustrating a configuration of an image processing circuit that implements a binarized average density error diffusion method. FIG. 11 is a diagram illustrating an example of an output image obtained by a binarized average density error diffusion method. FIG. 12 is a diagram illustrating an example of an output image obtained by a binarized average density error diffusion method. FIG. 13 is a diagram illustrating an example of an output image obtained by a binarized average density error diffusion method. FIG. 14 is a block diagram illustrating a configuration of an image processing circuit that implements a dither error diffusion method. FIG. 15 is a diagram showing an output example when output is performed by the dither error diffusion method. FIG. 16 is a diagram illustrating an example of output when output is performed by a dither error diffusion method. FIG. 17 is a diagram illustrating an output example when output is performed by a dither error diffusion method. FIG. 18 is a block diagram showing an image processing circuit when the binarized average error diffusion method and the dither error diffusion method are simply combined. FIG. 19 is a diagram showing an example of an output image when the binarized average error diffusion method and the dither error diffusion method are simply combined. FIG. 20 is a diagram illustrating an example of an output image when the binarized average error diffusion method and the dither error diffusion method are simply combined. FIG. 21 is a diagram illustrating an example of an output image when the binarized average error diffusion method and the dither error diffusion method are simply combined. [Description of Signs] 10 Image processing circuit 11 Multi-valued dither circuit 12 Error correction unit 13 Binarization unit 14 Binarization error calculation unit 15 Binary error diffusion unit 16 Binary average error diffusion unit 17 Attribute discrimination unit 18 Adjustment unit

Continuation of the front page (58) Field surveyed (Int. Cl. ⁷ , DB name) H04N 1/40-1/409 H04N 1/46 H04N 1/60

Claims (1)

(57) [Claim 1] N2 for converting input image data represented by an N1 value into image data having an N2 value smaller than the N1 value.
Value conversion means, N3 value conversion means for converting the image data represented by the N2 value into N3 value data smaller than the N2 value, and a first error for preparing data for diffusing an error in the N3 value conversion Diffusion data preparation means; and second error diffusion data for preparing data for diffusing an error between the average density of the output image at the time of the N3 value conversion and the N2 value image around the pixel of interest. Preparation means; and a station for the input image data based on the N2 value image data.
Difference, variance and edge detection between the maximum and minimum values of the area
And means for determining one of the amount, the maximum value of the local region according to the determination result of the previous SL Determination means
Means for adjusting the difference, the variance or the edge detection amount between the first error diffusion data and the minimum error value to be larger, so that the second error diffusion data is fed back larger; and the first error diffusion data and the adjusted second error diffusion data Means for correcting the N2 value image data by using the image processing apparatus.