JP2004096709A - Image processing apparatus and image processing method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To decrease a required buffer memory capacity while reducing the influence of a calculation error in error diffusion processing. <P>SOLUTION: A quantization circuit 4 quantizes input image data and outputs an output code. The quantization error which has occurred in the quantization circuit 4 is calculated by an inverse quantization circuit 5 and a subtractor 6. The calculated quantization error is stored in a buffer 8. Since the buffer 8 only needs to have a size capable of storing the quantization error, its size can be made smaller than before. A diffusion filter 9 diffuses the quantization error by using a quantization error or the like which is stored in the buffer 8. A latch 3 and a bit connector 1 are useful for reducing the influence of the calculation error in error diffusion processing on the next input image data. <P>COPYRIGHT: (C)2004,JPO

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to an image processing method and apparatus, and more particularly to an image processing method for quantizing input data into binary or multi-valued data while storing a difference between an input image density and an output image density by an error diffusion method or the like. And a device.
[0002]
[Prior art]
Conventionally, pseudo halftone processing is used to represent input multi-valued data as multi-valued data having a level lower than that of binary or input multi-valued data. As one example, an error diffusion method is known. This error diffusion method is described in "An Adaptive Algorithm for Spatial Gray Scale" in society for Information Display 1975 Symposium Digest of Technical, Technical Paper, 36, 1973. In this method, the pixel of interest is P, the density of the pixel is v, and the densities of the pixels P0, P1, P2, and P3 around the P before binarization processing are v0, v1, v2, and v3, respectively. Using a threshold value T for binarization as T, weighting the binarization error E at the point of interest P using weighting coefficients W0, W1, W2, and W3 empirically obtained, and surrounding pixels P0, P1, P2, It is allocated to each of P3. That is, this method is a method of making the average density of the output image macroscopically equal to the density of the input image. At this time, assuming that the output binary data is o, errors E0, E1, E2, and E3 with respect to the peripheral pixels P0, P1, P2, and P3 can be obtained by the following equations.
[0003]
If v ≧ T, o = 1, E = v−Vmax
If v <T, o = 0, E = v-Vmin
(However, Vmax: maximum density, Vmin: minimum density)
E0 = E × W0;
E1 = E × W1;
E2 = E × W2;
E3 = E × W3;
(Example of weight coefficient: W0 = 7/16, W1 = 1/16, W2 = 5/16, W3 = 3/16)
Incidentally, an error buffer is required to propagate this error. As can be seen from the above example, the weight coefficient is a decimal number of 1 or less.
[0004]
Therefore, assuming that the number of bits of the input image is 8 and the threshold value for binarization is fixed at 128 in the above example of the weighting coefficient, the quantization error is -127 to 127. There was a disadvantage that a bit was required.
[0005]
As a method for solving this, in Patent Literature 1 and Patent Literature 2, the sum of the error distribution values obtained by rounding down the decimal part of the result of multiplying the binarization error by the weighting coefficient is calculated, and the sum of the error distribution values is calculated. A method of calculating a residual error from a difference and redistributing it to peripheral pixels is disclosed.
[0006]
Further, in Patent Document 3, the lower n bits are extracted from the binarization error, and the binarization error obtained by complementing the lower n bits with 0 is multiplied by a weight coefficient to obtain an error distribution value. Is added to one of the above error distribution values and distributed to peripheral pixels.
[0007]
Patent Document 4 discloses a method in which the upper bits of the binarization error are stored in a buffer memory, and the lower bits of the binarization error are latched and added to the next pixel.
[0008]
[Patent Document 1]
Japanese Patent Publication No. 06-0666676
[Patent Document 2]
Japanese Patent Publication No. 07-022334
[Patent Document 3]
Japanese Patent Publication No. 07-093682
[Patent Document 4]
Japanese Patent Application Laid-Open No. 05-075863.
[0009]
[Problems to be solved by the invention]
However, the method disclosed in Patent Document 1 or Patent Document 2 has a drawback that a sum of error distribution values must be obtained and a residual error must be calculated from a difference from the binarization error.
[0010]
Further, in the method disclosed in Patent Document 3, the above-mentioned residual error does not have to be calculated, but the bit rate of the binarization error used in the calculation of the error distribution is reduced, and therefore, the distribution ratio is set to the magnitude of the binarization error. There is a disadvantage that it changes depending on. For example, in the case of the above-mentioned weighting coefficient, the lower 4 bits of the binarization error are multiplied by 0 to make the residual error 0, but when the binarization error is 15 or less, all become 0. Therefore, all binarization errors are distributed to a specific pixel.
[0011]
Further, in the method disclosed in Patent Document 4, since the upper bits of the binarization error are stored in the buffer memory, the buffer memory can be greatly reduced. Since the bit precision of the binarization error used for the calculation is reduced, there is a disadvantage that the distribution ratio changes depending on the magnitude of the binarization error. For example, in the case of the above-mentioned weight coefficient, in order to make the residual error zero, the upper four bits of the binarization error must be stored in the buffer memory, and the lower four bits must be stored in the latch. In the case of 15 or less, all become zero, so that all binarization errors are distributed to the next pixel. Further, in order to increase the accuracy of the error distribution, when the upper 5 bits or more of the binarization error is stored in the buffer memory, an error due to rounding (truncating) occurs. Further, Patent Document 4 describes that “the lower bits are stored in the latch circuit 4 after the sign is extended,” but the sign of the lower bits is extended when the binary error is negative. In this case, 1 must be added to the upper bit.
[0012]
For example, -1 is 11111111 in an 8-bit binary, but is divided into 1111 and 1111 when it is divided into upper and lower 4 bits. If the lower bit is negative-sign-extended, it becomes 11111 and represents -1. However, since the upper bit 1111 has a meaning of 11110000, it means -16. Therefore, the result of adding the upper and lower bits is -17. Therefore, in order to avoid a mismatch, it is necessary to add 1 to the upper bit 1111 to obtain 0000 (carry is not considered). Further, since the lower bits of the binarization error are added to the input pixel data, there is a disadvantage that the range of the binarization error is expanded and the number of bits is increased by one bit. For example, when latching the lower 4 bits of the binarization error, the normal binarization error has a range of -127 to 127 and can be represented by 8 bits, but the method disclosed in Japanese Patent Application Laid-Open No. 05-075863 is disclosed. In this case, the lower 4 bits of the binarization error (a range from -15 to 15 because of sign extension) are added to the input pixel data, so the range of the binarization error is -142 to 142, and 9 bits are required. Become.
[0013]
An object of the present invention is to solve the above-described problems, and an object of the present invention is to provide an image processing method and apparatus capable of reducing the capacity of a buffer memory while compensating for the influence of a calculation error at the time of error diffusion.
[0014]
[Means for Solving the Problems]
According to one aspect of the present invention, quantization means for quantizing input image data, calculation means for calculating a quantization error generated by a quantization process of the quantization means, An error for diffusing the quantization error based on at least the buffer to be stored, the quantization error of the first pixel stored in the buffer, and the quantization error of the second pixel calculated by the calculation unit. An image processing apparatus includes: a diffusion unit; and a reduction unit configured to reduce an influence of a calculation error by the error diffusion unit on subsequent input image data.
[0015]
As described above, according to the present invention, since the quantization error is stored in the buffer memory, the buffer memory only needs to have at least the number of bits corresponding to the quantization error, and the capacity is reduced as compared with the related art. It becomes possible. Further, by providing the reduction means, it is possible to eliminate the influence of the calculation error, and particularly to improve the image quality of the highlighted portion of the image.
[0016]
For example, the reduction unit may include a coupling unit that couples a decimal part of the correction value generated to diffuse the quantization error by the error diffusion unit to a lower bit side of the next input image data. That is, after the decimal part after the error diffusion operation is connected to the lower bit side of the input image data, the input image data is corrected, so that the output image can have high image quality.
[0017]
Further, if it is inappropriate to propagate the correction value to the next pixel and subsequent pixels, the image processing apparatus may further include a blocking unit that blocks the propagation of the correction value. Inappropriate cases include, for example, when the target pixel is the first pixel of the line, when the target pixel has a value of the lower limit level of the input image (eg, white pixel), or when the target pixel is the value of the upper limit level of the input image. (Eg, a black pixel). Of course, it goes without saying that the present invention can be applied to inappropriate cases other than the examples. Here, the lower limit level and the upper limit level may be not only one value but also a plurality of values existing within a predetermined range.
[0018]
Further, a holding means for holding a decimal part of the correction value, and when it is inappropriate to combine the decimal part of the correction value held in the holding means to the lower bit side of the next input image data, Clearing means for clearing the decimal part held by the holding means.
[0019]
Further, when the scanning direction of the input image is reversed, the image processing apparatus may further include a processing restricting unit that restricts the clearing process of the clearing unit.
[0020]
For example, if the scanning direction is reversed for each line, it is possible to limit the clearing process so that it is cleared to 0 only at the beginning of the image. Therefore, even when the scanning direction is reversed, an effect of high image quality can be expected.
[0021]
The image processing apparatus may further include a numerical value limiting unit that limits the quantization error calculated by the calculating unit to a numerical value within a predetermined range. As a result, the size of the buffer memory for storing the quantization error can be saved.
[0022]
According to another aspect of the present invention, a bit extension unit for bit-extending the image data of the pixel of interest, a correction unit for correcting the bit-extended image data, and quantizing an integer part of the corrected image data Quantization means, holding means for holding a quantization error generated by the quantization means, a first quantization error held by the holding means, and a second quantization error for the pixel of interest. Based on at least a correction value generation unit that generates a correction value used by the correction unit, and in the bit expansion process of the bit expansion unit, the correction value combined with the lower bit side of the next image data. An image processing apparatus comprising: a storage unit for storing a small number of parts.
[0023]
As described above, according to the present invention, since the quantization error is stored in the buffer memory, the buffer memory only needs to have at least the number of bits corresponding to the quantization error, and the capacity is reduced as compared with the related art. It becomes possible. Furthermore, after combining the decimal part after the error diffusion operation to the lower bit side of the input image data, the input image data is corrected, so that the quality of the output image can be improved.
[0024]
According to still another aspect of the present invention, a quantization unit for quantizing upper bits of input image data, and a calculation unit for calculating a quantization error generated by a quantization process of the quantization unit,
A buffer for storing the calculated quantization error, a quantization error of a first pixel stored in the buffer, and a quantization error of a second pixel calculated by the calculation unit. Error diffusion means for error-diffusing image data of a third pixel, holding means for holding a value of a predetermined bit or less of the error-diffused image data, and an integer part of the held value. Adding means for adding the fractional part of the held value to the lower-order bit side of the image data to which the integer part has been added, and outputting the result to the quantizing means. An image processing apparatus comprising:
[0025]
In this way, by storing the upper bits of the quantization error in the buffer memory, the capacity of the buffer memory can be reduced. Further, the lower bits of the integer part of the image data to which the correction value has been added by the error diffusion means are added to the next input image, and the decimal part of the correction value from the error diffusion means is connected to the lower bit side of the input image data. Then, since the input image data is configured to be corrected, the influence of the calculation error can be eliminated.
[0026]
Further, the calculation means may include a numerical limit means for limiting the calculated quantization error to a predetermined range and outputting the result to the buffer. Thus, the size of the buffer memory can be further reduced.
[0027]
Further, the maximum value of the representative quantization value used in calculating the quantization error may be set to be equal to or larger than the maximum value of the input image data. In this way, it is possible to prevent the quantization errors from accumulating without correction and causing the processing to fail.
[0028]
Further, the step width of the quantization representative value used in calculating the quantization error may be a constant value of a power of two. For example, as will be described in a second embodiment described later, if all the quantized representative values are multiples of 16, the output of the inverse quantizer can be 3 bits, so that the configuration of the calculating means can be simplified. it can.
[0029]
Further, first detecting means for detecting a pixel having a lower limit level value of the input image, and detecting a pixel having a lower limit level value of the input image as an output code relating to the pixel, the quantized representative value having a minimum value. And a first code output unit that outputs an output code such as:
[0030]
Further, a second detecting means for detecting a pixel having the upper limit level value of the input image, and detecting a pixel having the upper limit level value of the input image, the output representative code of the pixel has a maximum quantization representative value. And second code output means for outputting an output code such that
[0031]
Further, when a pixel having the lower limit level value or the upper limit level value is detected, a replacement unit that replaces a quantization error relating to the pixel with 0 may be provided.
[0032]
By doing so, the propagation of unnecessary errors can be suppressed, and as a result, the followability of edges can be improved, and furthermore, the loss of fine lines and the contamination of highlight portions can be prevented.
[0033]
According to still another aspect of the present invention, there is also provided an image processing method and an image processing program corresponding to the above-described image processing apparatus.
[0034]
BEST MODE FOR CARRYING OUT THE INVENTION
An embodiment of the present invention will be described below. Of course, the following embodiments are provided for the purpose of facilitating implementation by those skilled in the art of the present invention, and only a part of the scope of the present invention defined by the claims is set forth. It is only an embodiment of the section. Therefore, it will be obvious to those skilled in the art that even embodiments not directly described in the present specification are included in the technical scope of the present invention as long as the technical ideas are common.
[0035]
Although a plurality of embodiments will be described for the sake of convenience, those skilled in the art can easily understand that these inventions can be realized not only individually as inventions but also by appropriately combining a plurality of embodiments. I can do it.
[0036]
[First Embodiment]
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 is a block diagram illustrating the configuration of the image processing apparatus according to the first embodiment of the present invention.
[0037]
In FIG. 1, reference numeral 1 denotes a bit coupling circuit which couples the decimal accumulated value from the latch 3 to the lower order of the input image data. An adder 2 adds the correction value from the diffusion filter 9 to the input image data. A latch 3 delays the decimal part of the adder 2 by one pixel. Numeral 4 denotes a quantizer which compares the integer part of the adder 2 with a threshold and converts it into an output code. Reference numeral 5 denotes an inverse quantizer, which generates a quantized representative value from an output code. Reference numeral 6 denotes a subtractor, which generates a quantization error by subtracting a quantization representative value from a value immediately before quantization. A limiter 7 limits a quantization error from the subtractor 6 to a predetermined range (existing range). A line buffer 8 delays the quantization error from the limiter 7 by about one line. Reference numeral 9 denotes a diffusion filter, which multiplies a coefficient corresponding to a quantization error of a current pixel and a diffusion target pixel of the previous line from the line buffer 8 and adds the sum to the next input image data by the adder 2.
[0038]
Next, a processing flow will be described. Hereinafter, specific numerical values when the present invention is applied to five-value error diffusion will be described. In the present embodiment, in order to reduce the capacity of the line buffer 8, instead of storing the value after error diffusion (output value from the diffusion filter) in the line buffer, the quantization error itself is stored in the line buffer. I have to. Since the value of the quantization error (eg, 6 bits) is smaller than the value after error diffusion (eg, 10 bits), the buffer capacity can be saved.
[0039]
FIG. 2 shows an example of the diffusion coefficient used for the operation of the diffusion filter 9. As shown in FIG. 2, in the method of obtaining the correction value of the next pixel from the surrounding quantization error (minimum average error method), the diffusion coefficient and the position of the error diffusion method have a point-symmetric relationship.
[0040]
FIG. 1 shows the number of bits of each signal line when the input image data is 8 bits and quinary error diffusion is performed using the diffusion coefficient of FIG. Since the denominator of the diffusion coefficient in FIG. 2 is 16, the sum of the correction value from the diffusion filter 9 and the output image data from the bit combination circuit 1 is 4 bits for the decimal part and an integer part (-31 to 286). To a total of 14 bits. The fractional part of the corrected image data output from the adder 2 is delayed by one pixel in the latch 4 and input to the adder 2 together with the next pixel. Therefore, the decimal part due to the diffusion coefficient is propagated to the next pixel and beyond without being truncated, so that the quantization error is corrected (all of the quantization error is propagated).
[0041]
On the other hand, the integer part of the corrected image data is compared with a predetermined threshold by the quantizer 4 and converted into a quinary code. Here, assuming that the input image data is x, the output code c and the representative quantization value r are as follows.
[0042]
When x <32, c = 0, r = 0
C = 1, r = 64 when 32 ≦ x <96
When 96 ≦ x <160, c = 2, r = 128
When 160 ≦ x <224, c = 3, r = 192
When 224 ≦ x, c = 4, r = 255.
[0043]
Therefore, when the threshold value of the quantizer is set at the center of each of the representative quantization values, the quantization error is in the range of −31 to 31. On the other hand, since the correction value due to the quantization error of the surrounding pixels is added to the input image data, the output of the adder 6 exceeds the range of −31 to 31, but out of the range of −31 to 31. The case where the corrected image data exceeds the existing range of the input image data. Therefore, there is no problem even if the calculated quantization error is numerically limited to a predetermined range by the limiter 7.
[0044]
Since the quantization error is limited by the limiter 7, it is sufficient that the line buffer 8 stores a quantization error in the range of -311 to 31. Therefore, the number of bits of the line buffer 8 may be six. This means that since the output of the diffusion filter 8 is 10 bits, the size can be saved by 4 bits.
[0045]
The diffusion filter 8 multiplies the current pixel and the quantization error read out from the line buffer 8 one line before by the diffusion coefficient in FIG. 2 to obtain a correction value, and outputs the correction value to the adder 2.
[0046]
With the above processing, the quinary processing for one input image data is completed. If the processing has not been completed for all the pixels, the above processing is repeated by shifting one pixel until the processing is completed. By this repetition processing, the quinary processing for the entire image is completed.
[0047]
Incidentally, the latch 3 is cleared to 0 at the head of the line in the case of a normal raster scan. That is, the last pixel of the previous line and the first pixel of the current line are pixels that are not physically adjacent to each other, and are pixels having low correlation of image data, but are adjacent to each other in processing. Since the number of pixels is one, it is cleared to 0 at the head of the line in order to suppress the propagation of unnecessary errors. On the other hand, in the case of the scanning method in which the scanning direction is inverted for each line, the two pixels are physically and continuously continuous even at the time of scanning inversion (when shifting from the previous line to the current line). Therefore, it is not necessary to clear the error to 0 at the head of the line, and it is preferable to propagate the error from the viewpoint of image quality. Therefore, when inverting the scanning direction, it is preferable to prohibit the clearing of 0 at the head of the line, and to perform the clearing only at the head of the image or at the head of the line when the scanning direction is not inverted.
[0048]
As described above, according to the first embodiment of the present invention, since the quantization error is stored in the buffer while being limited to the predetermined range, the capacity of the buffer can be reduced. In addition, the rounding error of the error diffusion operation can be corrected by bit-combining the decimal part at the time of the error diffusion operation to the lower order of the next image data, and as a result, the image quality is improved, and in particular, the highlighting with a small quantization error is performed. It is possible to improve the image quality of the section. Also, the effect of buffer reduction will increase as the number of levels after error diffusion increases.
[0049]
[Second embodiment]
FIG. 3 is a block diagram illustrating a configuration of an image processing apparatus according to the second embodiment of the present invention. In the figure, 11 and 12 are adders, 13 is a latch, 14 is a quantizer, 15 is an inverse quantizer, 16 is a subtractor, 17 is a limiter, 18 line buffers, and 19 is a diffusion filter. Hereinafter, only portions different from the first embodiment will be described.
[0050]
In the present embodiment, the capacity of the line buffer 18 is reduced by extending the number of bits of the latch 13 by m bits and accumulating not only the decimal part but also the lower m bits of the integer part after error correction.
[0051]
Accordingly, the decimal part of the output of the latch 13 is coupled to the lower part of the input image data as in the first embodiment, but the m-bit integer part is added to the input image by the adder 11. Therefore, the output after the bit combination is 13 bits. The 10-m bits of the separated integer part are converted into an output code by the quantizer 14, become a representative quantization value by the inverse quantizer 15, and a quantization error is generated by the subtracter 16. . This quantization error is limited to a significant value by the limiter 17 as in the first embodiment, and is delayed by about one line in the line buffer 18. The quantization error of the current pixel and the quantization error of the adjacent previous line are input to the diffusion filter 19, and the diffusion coefficients corresponding to each are multiplied and summed, and the correction value of the next pixel is calculated. The calculated correction value is added by the adder 12 to the next pixel.
[0052]
Here, for convenience of explanation, if m = 2, the latch 13 has 6 bits, and the output range of the adder 11 is 0 to 258. On the other hand, the input x of the quantizer 14 is 8 bits (−7 to 72), and the output code c and the quantized representative value r (the actual weight is shifted left by 2 bits, that is, quadrupled) as follows: Is required.
[0053]
When x <8, c = 0, r = 0
C = 1, r = 16 when 8 ≦ x <24
When 24 ≦ x <40, c = 2, r = 32
When 40 ≦ x <56, c = 3, r = 48
When 56 ≦ x, c = 4 and r = 64.
[0054]
Here, the reason for r = 64 when 56 ≦ x is as follows. That is, when r = 63, 63 × 4 = 252 in actual input value conversion (2-bit left shift), which is less than the input image range, and as a result, 255 input data are continuous. In this case, the quantization errors of 255-252 = 3 may be accumulated without correction, and may exceed the output range of the adder 12 and break down. Therefore, in order to suppress such a failure of the processing, it is necessary to set the maximum value of the quantization representative value to a value equal to or larger than the maximum value of the input image data.
[0055]
Although this operation slightly lowers the density of the solid black portion (1/256 level), there is almost no effect on the vicinity of the solid black because the density originally changes little. Therefore, the range of the significant quantization error is -8 to 7, and the line buffer 20 is reduced to 4 bits.
[0056]
In the present embodiment, the quantization representative values are all multiples of 16 so as to have a step width of a power of two. As a result, the output of the inverse quantizer 15 may be 3 bits, and the inverse quantizer 15 and the subtractor 16 are simplified.
[0057]
In the present embodiment, where the quantization error is small, such as when the input pixel data is close to the quantization representative value, the error distribution to the next pixel is relatively large. In particular, when the quantization error is 3 or less, 100% of the error is distributed to the next pixel. This hastens the generation of dots in parts with a small quantization error, such as highlights, and has the effect of improving the delay in dot generation during error diffusion in parts with large changes in density, called `` sweeping ''. is there.
[0058]
On the other hand, when the input data is farther than the representative quantization value, the quantization error becomes large, and the distribution ratio of the error is almost the same as in the conventional case, and the dispersibility of the dots is maintained. does not change).
[0059]
Even if the input data is near the quantization representative value, if the next quantization representative value (farther than the input data) is selected by accumulating the error, the quantization error becomes large. Is almost the same as before. Therefore, if m is not too large, the dispersibility of the dots is maintained (the texture due to the connection of the dots is not much different from the conventional one).
[0060]
When the texture due to the connection of dots becomes conspicuous because m is large, the image quality can be improved by changing the diffusion coefficient. That is, when the quantization error of the immediately preceding pixel is 0, the diffusion coefficient immediately below is increased as shown in FIG. This compensates for the proportion of quantization error propagated downward, and improves the dispersibility of the dots in the vertical direction. Note that the condition for changing the diffusion coefficient is limited to the case where the quantization error of the immediately preceding pixel is 0, because the substantial distribution ratio when the quantization error is 0 is particularly biased (quantum The variation of the substantial distribution ratio is relatively small except when the conversion error is 0). Further, since the substantial distribution ratio also depends on the input data, a similar effect can be obtained by changing the diffusion coefficient according to the level of the input data.
[0061]
[Third embodiment]
FIG. 4 is a block diagram illustrating a configuration of an image processing apparatus according to the third embodiment of the present invention. In the figure, 20 is a white pixel / black pixel detector, 21 is a latch, 22 is a code replacement circuit, and 23 is a quantization error replacement circuit. Hereinafter, only portions different from the second embodiment will be described.
[0062]
The white pixel / black pixel detector 20 detects white pixels (0 in the second embodiment) and black pixels (255 in the second embodiment) in the input image data. The detection result is output to the latch 21 and the replacement circuits 22 and 23.
[0063]
When the white pixel / black pixel detector 20 detects a white pixel or a black pixel, the latch 21 clears the latch and prevents propagation of an error to the next pixel and subsequent pixels. As a result, it is possible to suppress inappropriate results such as contamination of a highlight portion and disappearance of a thin line due to propagation of an unnecessary error.
[0064]
The code replacement circuit 22 outputs a code indicating a white pixel (0 in the second embodiment) when a white pixel is detected, and outputs a code indicating a black pixel (the second code) when a black pixel is detected. In the embodiment, 4) is output, and otherwise, the input code is output.
[0065]
The quantization error replacement circuit 23 outputs 0 when a white pixel and a black pixel are detected, and otherwise outputs a quantization error limited to a significant value as in the above-described embodiment.
[0066]
Although the effect can be obtained by simply providing at least one of the clear processing of the latch 21 and the code replacement circuit 22 and the quantization error replacement circuit 23, two of these may be employed in combination or all three may be employed. May be adopted. A higher effect can be expected by adopting more.
[0067]
In the present embodiment, the values of the upper limit and the lower limit of the input image data are detected, and when the values of the upper limit and the lower limit of the input image data are set, the quantization error is set to 0, and the error is not propagated to the next pixel and subsequent pixels. It is. With this configuration, it is possible to suppress problems such as a decrease in density of a solid black portion, disappearance of a thin line due to a diffusion error, and generation of a dot in a white pixel portion as described in the second embodiment.
[0068]
In this embodiment, white pixels and black pixels are detected, but those skilled in the art can understand that this is an extreme example. The present invention is not limited to this. For example, instead of detecting a white pixel and a black pixel, a pixel below a predetermined level or a pixel above a predetermined level may be detected and the above processing may be performed. The predetermined level is a so-called threshold value, and the threshold value can be appropriately determined by a trade-off with image deterioration. For example, a pixel at a level that can be regarded as a substantially white pixel can be substantially used as a white pixel by this threshold value. (However, if the threshold is increased, deterioration such as overexposure and underexposure of black will become severe. Therefore, it is usually desirable to set the maximum value and the minimum value of the input image data.)
[Other embodiments]
A storage medium storing software program codes for realizing the functions of the above-described embodiments is supplied to a system or an apparatus, and a computer (or CPU or MPU) of the system or the apparatus executes the program code stored in the storage medium. It goes without saying that the object of the present invention is also achieved by executing the reading.
[0069]
In this case, the program code itself read from the storage medium realizes the novel function of the present invention, and the storage medium storing the program code constitutes the present invention.
[0070]
As a storage medium for supplying the program code, for example, a flexible disk, a hard disk, an optical disk, a magneto-optical disk, a CD-ROM, a CD-R, a magnetic tape, a nonvolatile memory card, a ROM, and the like can be used.
[0071]
The functions of the above-described embodiments are implemented when the computer executes the readout program code, and the OS or the like running on the computer performs part of the actual processing based on the instructions of the program code. Alternatively, all the operations are performed, and the functions of the above-described embodiments can be realized by the processing.
[0072]
Further, after the program code read from the storage medium is written to a memory provided in a function expansion board inserted into the computer or a function expansion unit connected to the computer, the function expansion is performed based on the instruction of the program code. The CPU or the like provided in the board or the function expansion unit performs part or all of the actual processing, and the functions of the above-described embodiments can also be realized by the processing.
[0073]
The invention of the present application is also applicable to a case where the program is delivered to a requester via a communication line such as a personal computer communication from a storage medium storing a program code of software for realizing the functions of the above-described embodiments. It goes without saying that it can be applied.
[0074]
【The invention's effect】
As described above, according to the first aspect, since the image processing apparatus is configured to store the quantization error in the buffer memory, the buffer memory need only have the number of bits corresponding to the quantization error. The size of the buffer memory can be reduced more than before. In addition, by combining the decimal part after the error diffusion calculation with the lower bit side of the input image data and correcting the input image data, the influence of the calculation error can be reduced. In particular, higher image quality can be expected in the highlight portion of the image.
[0075]
Further, according to the second aspect, the image processing apparatus is configured to store only the upper bits of the quantization error in the buffer memory, so that the size of the buffer memory can be made smaller than in the first aspect. In addition, if a configuration in which a predetermined bit or less of the image data obtained by adding the correction values output from the error diffusion filter is cumulatively added, the influence of the calculation error can be reduced.
[0076]
According to the third aspect, when the target pixel is a white pixel (lower limit of the input image range) and a black pixel (upper limit of the input image range), the propagation of unnecessary errors is suppressed. The followability can be improved, and the loss of fine lines and the stain of highlights can be prevented.
[Brief description of the drawings]
FIG. 1 is a block diagram illustrating a configuration of an exemplary image processing apparatus according to a first embodiment;
FIG. 2 is a diagram showing an example of a position of a quantization error and a diffusion coefficient used for an operation of a diffusion filter 9;
FIG. 3 is a block diagram illustrating a configuration of an exemplary image processing apparatus according to a second embodiment;
FIG. 4 is a block diagram illustrating a configuration of an exemplary image processing apparatus according to a third embodiment;
FIG. 5 is a diagram illustrating an example of a position of a quantization error and a diffusion coefficient used for the operation of the diffusion filter 19 when the quantization error of the immediately preceding pixel is 0.
[Explanation of symbols]
1 ... bit coupling circuit
2 ... Adder
3 ... Latch
4: Quantizer
5 Inverse quantizer
6 ... Subtractor
7 ... Limiter
8 ... Line buffer
9 ... Diffusion filter
11 ... Adder
12 ... Adder
13 ... Latch
14 Quantizer
15 ... Inverse quantizer
16 ... Subtractor
17 ... Limiter
18 ... Line buffer
19 ... Diffusion filter
20 white / black pixel detector
21 ... Latch
22 Code replacement circuit
23 ... Quantization error replacement circuit

Claims (49)

Quantization means for quantizing the input image data;
Calculation means for calculating a quantization error generated by the quantization processing of the quantization means,
A buffer for storing the calculated quantization error,
Error diffusion means for diffusing the quantization error based at least on the quantization error of the first pixel stored in the buffer and the quantization error of the second pixel calculated by the calculation means;
An image processing apparatus comprising: a reduction unit configured to reduce an influence of a calculation error by the error diffusion unit on next input image data. 2. The image according to claim 1, wherein the reducing unit further includes a combining unit that combines a decimal part of the correction value generated to diffuse the quantization error by the error diffusion unit to a lower bit side of the next input image data. Processing equipment. The image processing apparatus according to claim 2, further comprising a blocking unit that blocks propagation of the correction value when it is inappropriate to propagate the correction value to the next pixel and subsequent pixels. Holding means for holding a decimal part of the correction value,
When it is inappropriate to combine the decimal part of the correction value held by the holding means with the lower bit side of the next input image data, clear means for clearing the decimal part held by the holding means ,
The image processing apparatus according to claim 2, further comprising: The image processing apparatus according to claim 4, further comprising a processing restriction unit that restricts a clear process of the clear unit when a scanning direction of the input image is reversed. The inappropriate case is that at least one of a case where the pixel of interest is the first pixel of the line, a case where the pixel of interest has a value of the lower limit level of the input image, and a case where the pixel of interest has a value of the upper limit level of the input image. The image processing apparatus according to claim 3, wherein the image processing apparatus includes the image processing apparatus. The image processing apparatus according to claim 1, further comprising a numerical value limiting unit that limits the quantization error calculated by the calculating unit to a numerical value within a predetermined range. Bit extension means for bit-extending the image data of the pixel of interest,
Correction means for correcting the bit-extended image data,
Quantizing means for quantizing an integer part of the corrected image data,
Holding means for holding a quantization error generated by the quantization means,
A correction value generation unit configured to generate a correction value used by the correction unit based at least on the first quantization error held by the storage unit and the second quantization error related to the pixel of interest;
In the bit expansion processing of the bit expansion means, storage means for storing a decimal part of the correction value to be combined with the lower bit side of the next image data,
An image processing apparatus comprising: Quantizing means for quantizing upper bits of the input image data,
Calculation means for calculating a quantization error generated by the quantization processing of the quantization means,
A buffer for storing the calculated quantization error,
An error for error-diffusion of image data of a third pixel based at least on the quantization error of the first pixel stored in the buffer and the quantization error of the second pixel calculated by the calculation unit. Spreading means;
Holding means for holding a value equal to or less than a predetermined bit of the error-diffused image data, and adding means for adding an integer part of the held value to input image data,
Bit combining means for combining the fractional part of the held value with the lower bit side of the image data to which the integer part has been added, and outputting the combined value to the quantization means,
An image processing apparatus comprising: The image processing apparatus according to claim 9, wherein the calculating unit includes a numerical value limiting unit that limits the calculated quantization error to a predetermined range and outputs the result to the buffer. The image processing according to claim 9, wherein a maximum value of a quantization representative value used when calculating the quantization error is set to be equal to or larger than a maximum value of input image data. apparatus. 12. The image processing apparatus according to claim 11, wherein a step width of a quantization representative value used when calculating the quantization error is set to a constant value of a power of two. 10. The image processing apparatus according to claim 9, further comprising: a blocking unit configured to block the propagation of the held value when it is inappropriate to propagate the held value to the next pixel and subsequent pixels. The image processing device according to any one of claims 12 to 12. The blocking means,
14. The image processing apparatus according to claim 13, further comprising a clearing unit that clears the value held in the holding unit when the inappropriate state is found. 15. The image processing apparatus according to claim 14, further comprising a processing restriction unit that restricts a clear process of the clear unit when a scanning direction of the input image is reversed. The inappropriate case is that at least one of a case where the pixel of interest is the first pixel of the line, a case where the pixel of interest has a value of the lower limit level of the input image, and a case where the pixel of interest has a value of the upper limit level of the input image. The image processing apparatus according to claim 13, wherein the image processing apparatus includes the image processing apparatus. First detection means for detecting a pixel having a lower limit level value of the input image;
When a pixel having a lower limit level value of the input image is detected, a first code output unit that outputs an output code having a minimum quantization representative value as an output code for the pixel,
The image processing apparatus according to any one of claims 9 to 16, further comprising: Second detection means for detecting a pixel having an upper limit level value of the input image;
When a pixel having an upper limit level value of the input image is detected, as an output code related to the pixel, a second code output unit that outputs an output code that maximizes a quantization representative value,
The image processing apparatus according to any one of claims 9 to 16, further comprising: 19. The image processing apparatus according to claim 9, further comprising: a replacement unit configured to replace a quantization error related to the pixel with 0 when a pixel having the lower limit level value or the upper limit level value is detected. . Quantizing the input image data;
Calculating a quantization error generated by the quantization;
Storing the calculated quantization error in a buffer;
Performing error diffusion based at least on the quantization error of the first pixel stored in the buffer and the calculated quantization error of the second pixel;
Reducing the influence of the calculation error due to the error diffusion on the next input image data. 21. The image processing according to claim 20, wherein the step of reducing includes a step of combining a decimal part of a correction value generated to diffuse a quantization error by the error diffusion to a lower bit side of next input image data. Method. 22. The image processing method according to claim 21, further comprising a step of preventing propagation of the correction value when it is inappropriate to propagate the correction value to a subsequent pixel. Holding a decimal part of the correction value;
When it is inappropriate to combine the decimal part of the held correction value with the lower bit side of the next input image data, clearing the decimal part held in the holding step,
The image processing method according to claim 21, further comprising: 24. The image processing method according to claim 23, further comprising a step of restricting the clearing process when the scanning direction of the input image is reversed. The inappropriate case is that at least one of a case where the pixel of interest is the first pixel of the line, a case where the pixel of interest has a value of the lower limit level of the input image, and a case where the pixel of interest has a value of the upper limit level of the input image. The image processing method according to any one of claims 22 to 24, wherein the image processing method is included. 26. The image processing method according to claim 20, further comprising a step of limiting the quantization error calculated in the calculation step to a numerical value within a predetermined range. Bit expanding the image data of the pixel of interest;
Correcting the bit-extended image data;
Quantizing an integer part of the corrected image data;
Holding a quantization error generated by the quantization,
Generating a correction value used in the correction step based at least on the held first quantization error and the second quantization error related to the pixel of interest;
Storing the fractional part of the correction value that is combined with the lower bit side of the next image data in the step of bit extension;
An image processing method comprising: Quantizing the upper bits of the input image data;
Calculating a quantization error generated by the quantization process;
Storing the calculated quantization error in a buffer;
Error-diffusion of image data of a third pixel based at least on the quantization error of the first pixel stored in the buffer and the calculated quantization error of the second pixel;
Holding the value of a predetermined bit or less of the error-diffused image data, and adding an integer part of the held value to the input image data,
Combining the fractional part of the held value with the lower bit side of the image data to which the integer part has been added, and outputting the combined value;
An image processing method comprising: 29. The image processing method according to claim 28, wherein the calculating step includes a step of limiting the calculated quantization error to a predetermined range and outputting the result to the buffer. 30. The image processing according to claim 28, wherein a maximum value of a quantization representative value used in calculating the quantization error is set to be equal to or larger than a maximum value of input image data. Method. 31. The image processing method according to claim 30, wherein a step width of a quantization representative value used in calculating the quantization error is a constant value of a power of two. 32. The method according to claim 28, further comprising, if it is inappropriate to propagate the retained value to the next pixel and subsequent pixels, blocking the propagation of the retained value. The image processing method according to claim 1. 33. The image processing method according to claim 32, wherein the step of blocking includes a step of clearing the value held in the holding step in the case of the inappropriateness. The image processing method according to claim 33, further comprising a step of restricting the clearing process when the scanning direction of the input image is reversed. The inappropriate case is that at least one of a case where the pixel of interest is the first pixel of the line, a case where the pixel of interest has a value of the lower limit level of the input image, and a case where the pixel of interest has a value of the upper limit level of the input image. The image processing method according to any one of claims 32 to 34, wherein the method is included. Detecting a pixel having a lower level value of the input image;
When a pixel having a lower limit level value of the input image is detected, outputting an output code such that a quantization representative value is minimized as an output code for the pixel,
The image processing method according to any one of claims 28 to 35, further comprising: Detecting pixels having an upper level value of the input image;
When a pixel having an upper limit level value of the input image is detected, outputting an output code such that a quantization representative value is maximized as an output code related to the pixel,
The image processing method according to any one of claims 28 to 35, further comprising: 38. The image processing method according to claim 28, further comprising: when a pixel having the lower limit level value or the upper limit level value is detected, replacing a quantization error relating to the pixel with 0. . Quantizing the input image data;
Calculating a quantization error generated by the quantization;
Storing the calculated quantization error in a buffer;
Performing error diffusion based at least on the quantization error of the first pixel stored in the buffer and the calculated quantization error of the second pixel;
A step of reducing the influence of the calculation error due to the error diffusion on the next input image data. Bit expanding the image data of the pixel of interest;
Correcting the bit-extended image data;
Quantizing an integer part of the corrected image data;
Holding a quantization error generated by the quantization,
Generating a correction value used in the correction step based at least on the held first quantization error and the second quantization error related to the pixel of interest;
Storing the fractional part of the correction value that is combined with the lower bit side of the next image data in the step of bit extension;
Image processing program for causing a computer to execute. Quantizing the upper bits of the input image data;
Calculating a quantization error generated by the quantization process;
Storing the calculated quantization error in a buffer;
Error-diffusion of image data of a third pixel based at least on the quantization error of the first pixel stored in the buffer and the calculated quantization error of the second pixel;
Holding the value of a predetermined bit or less of the error-diffused image data, and adding an integer part of the held value to the input image data,
Combining the fractional part of the held value with the lower bit side of the image data to which the integer part has been added, and outputting the combined value;
Image processing program for causing a computer to execute. Means for detecting that the quantization error of the immediately preceding pixel is 0;
20. The image processing apparatus according to claim 9, further comprising: a change unit configured to change an error diffusion coefficient for the pixel when the quantization error of the immediately preceding pixel is detected to be 0. . 43. The image processing apparatus according to claim 42, wherein the changing unit changes the ratio of a diffusion coefficient of an upper line to be larger. Means for calculating the difference between the input image data and the quantized representative value closest to the input image data,
20. The image processing apparatus according to claim 9, further comprising a changing unit configured to change an error diffusion coefficient according to the difference. The image processing apparatus according to claim 44, wherein the changing unit changes the ratio of the diffusion coefficient of the upper line to increase as the difference decreases. Detecting that the quantization error of the immediately preceding pixel is 0;
39. The image processing method according to claim 28, further comprising: when the quantization error of the immediately preceding pixel is detected to be 0, changing an error diffusion coefficient for the pixel. 47. The image processing apparatus according to claim 46, wherein the changing step changes the ratio of a diffusion coefficient of an upper line to be larger. Calculating a difference between the input image data and the quantized representative value closest to the input image data;
39. The image processing apparatus according to claim 28, further comprising: changing an error diffusion coefficient according to the difference. 49. The image processing method according to claim 48, wherein in the changing step, the ratio of the diffusion coefficient of the upper line increases as the difference decreases.

Images