JP3697027B2 - Image processing apparatus and method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an image processing apparatus and method, and more particularly, to an image processing apparatus and method for performing quantization by performing error diffusion processing on multivalued image data.
[0002]
[Prior art]
There is an error diffusion method as one of pseudo halftone processing for expressing a grayscale image. The error diffusion method binarizes each pixel of a grayscale image input in the order of raster scanning with a certain threshold value, and the difference value between the binarized value and the original value of the pixel (hereinafter “binarization”). In other words, a pseudo halftone image with high image quality is obtained by diffusing the difference value) to the peripheral pixels of the pixel.
[0003]
1A to 1D are diagrams for explaining the error diffusion method, where the number of gradations of the input grayscale image data is 256, the value representing white is “0”, the value representing black is “255”, The binarization threshold is set to “128”.
[0004]
Assuming that the pixel B shown in FIG. 1A is the pixel of interest and its value is “80”, for example, the pixel B is binarized to “0” representing white. The binarized difference value “80” between the original value “80” and the binarized value “0” is, for example, the ratio shown in FIG. 1B to the peripheral pixels C, F, and G of the pixel B. Diffused. That is, 80 × 1/4, 80 × 2/4, and 80 × 1/4 are added to the respective values of the pixels C, F, and G, and these are made new values of the pixels C, F, and G.
[0005]
For example, when the value of the pixel B is “140”, the pixel B is binarized to “255” representing black, and the binarized difference value “−115 (= 140−255)” is the pixel. It is diffused to the peripheral pixels of B. That is, −155 × 1/4, −155 × 2/4, and −155 × 1/4 are added to the respective values of the pixels C, F, and G, and these are made new values of the pixels C, F, and G.
[0006]
By performing the above processing on all the pixels input in the raster scan order, the input grayscale image can be converted into a pseudo halftone image.
[0007]
By the way, in the example of FIGS. 1A and 1B, if the binarized difference value generated by the binarization processing of the pixel X is represented by ΔX, the value αG of the pixel G in which the error is diffused is represented by the following equation.
αG = G + (δB × 1/4 + δC × 2/4 + δF × 1/4) (1)
[0008]
Among the added values, δB × 1/4 + δC × 2/4 (= βG) is the time required for the binarization processing for one line after the binarization processing of the pixel C, that is, the pixel G It is necessary to memorize the binarization process. In a facsimile machine or the like, as shown in FIG. 1C, this is realized by preparing a FIFO memory capable of holding information for one line of an input image.
[0009]
That is, when the binarization process for the pixel C is finished, βG used in the binarization process for the pixel G of the next line is saved in the FIFO memory as shown in FIG. As shown in FIG.1D by (b), the content of the FIFO memory is shifted by one cell every time the binarization process of one pixel is performed, and when the pixel G after one line of the pixel C becomes the target pixel, As shown by (c) in FIG. 1D, βG is output from the FIFO memory. Note that ΔF × 1/4 only needs to be stored for the time required for the binarization processing of one pixel, and therefore may be stored in a register or the like instead of the FIFO memory. Then, after performing the calculation of equation (1), binarization of αG is performed to obtain a new pixel value, and then binarization processing of the pixel H is performed.
[0010]
[Problems to be solved by the invention]
However, the above-described technique has the following problems. In other words, in order to perform the above error diffusion processing, the number of bits required to represent the gradation of the input image is n (n = 8 for 256 gradations), the number of pixels for one line is N, When an error diffusion matrix such as that shown in 1B is used, the same n bits as the input image are required to express βX, so a FIFO memory with a relatively large storage capacity of n × N bits is required.
[0011]
The present invention is to solve the above-described problem, and when performing quantization by error diffusion processing, the image quality of the output image is prevented and the storage capacity of the memory for storing the quantization error is saved. An object of the present invention is to provide an image processing apparatus and a method thereof.
[0012]
[Means for Solving the Problems]
The present invention has the following configuration as one means for achieving the above object.
[0013]
An image processing apparatus according to the present invention is an image processing apparatus that performs error diffusion processing on input multi-valued image data, and includes a setting unit that sets a threshold value and a constant according to the set number of output gradations, and a pixel Quantize the data with the threshold value, obtain a quantization error, and perform a predetermined operation on a plurality of quantization errors generated by the quantization in a certain line, and the predetermined pixel of the next line of the line A first computing means for computing an error integrated value to be added to the data, dividing the error integrated value by the constant and storing it in the storage means; and multiplying the data read from the storage means by the constant to calculate the predetermined pixel. And second quantizing means for adding to the data, wherein the quantizing means comprises quantizing the data of the pixels subjected to the calculation by the second calculating means.
[0014]
An image processing method according to the present invention is an image processing method for performing error diffusion processing on input multi-valued image data, and a setting step for setting a threshold value and a constant according to the set number of output gradations. A quantization step for quantizing the pixel data with the threshold value, obtaining a quantization error thereof, a first calculation step for dividing the quantization error generated by the quantization by the constant and storing it in the memory, and from the memory A second calculation step of multiplying the read data by the constant, performing a predetermined calculation and adding the result to the pixel data, and the quantization step of the pixel subjected to the calculation by the second calculation step. It is characterized by quantizing data.
[0015]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, an image processing apparatus according to an embodiment of the present invention will be described in detail with reference to the drawings.
[0016]
[Double error diffusion processing method]
When performing quantization by error diffusion processing, there are the following methods that prevent image quality deterioration of the output image and save the storage capacity of the FIFO memory that stores the quantization error.
[0017]
In the error diffusion processing method described above, βX is stored in the FIFO memory, but β′X obtained by shifting βX to the right by L bits (L is a predetermined value, 0 <L <n) is stored in the FIFO memory. Then, the carry value (remainder) generated as a result of the right shift is held until the processing of the pixel of interest Y next to the pixel X. A process of repeating βY obtained by adding a carry value to βY obtained at the time of processing of the pixel Y as βY, and storing β′Y right-shifted by L bits in the FIFO memory is repeated. When data is extracted from the FIFO memory, the extracted value is left-shifted by L bits, and the result is used for subsequent arithmetic processing.
[0018]
Thus, if βX is shifted L bits to the right and stored in the FIFO memory, the number of bits of each cell in the FIFO memory can be reduced by L bits. Therefore, the storage capacity of L × N bits can be saved in the entire FIFO memory.
[0019]
This method performs multi-valued, 2 ^ (nL) -value error diffusion processing on a number sequence βA, βB, βC,..., And thereby a sequence β′A, β′B, β′C, Since it can be considered that the error diffusion processing is performed on the image data, it is called a “double error diffusion processing method”. It has been experimentally confirmed that this method has almost no deterioration in image quality as compared with a normal error diffusion processing method. The above 2 ^ (nL) represents 2 to the power of (nL), and hereinafter, the power of a is referred to as “a ^ b”.
[0020]
Now, there are ink jet printers that can perform multi-value recording, that is, not only on / off of one dot unit but also one that can express ternary or more gradations in one dot. For example, an ink jet printer is configured to record dots in which either dark ink or light ink is selected. In this case, it is possible to record in three gradations: ink is not applied, light ink is applied, and dark ink is applied.
[0021]
The error diffusion method can also be used for such a recording apparatus. This is because this method allows not only binarization of the input image but also quantization of arbitrary steps such as ternary and quaternary values.
[0022]
For example, suppose that there is a recording device that can record each pixel with 5 gradations of density 0%, 25%, 50%, 75% and 100%. An example will be described. Assuming that the number of gradations of the input image is 256 (the gradation value representing white is 0, and the gradation value representing black is 255), this recording device uses gradations corresponding to gradation values 0, 64, 128, 192, and 255. It will be possible to record. Therefore, the input image is converted into five values using 32, 96, 160, and 224, which are in the middle of recordable gradation values, as threshold values, and an output value of 0 to 4 corresponding to the gradation value is assigned to each input pixel. Then, the recording apparatus records dots of density 0%, 25%, 50%, 75% and 100% according to the output value. At that time, the difference between the gradation value corresponding to the output value and the gradation value of the target pixel (quantization difference value) is set to a predetermined ratio to the peripheral pixels of the target pixel, as in the binarization process described above. Just diffuse.
[0023]
Some recording apparatuses can switch the number of output gradations. In such a recording apparatus, it is possible to freely select, for example, a binary, ternary, or quinary gradation, and when the binary is selected, the density is 0% or 100%. When a value is selected, dots of density 0%, 25%, 50%, 75% and 100% are formed and an image is recorded. In error diffusion processing corresponding to such a recording apparatus, it is essential to be able to change processing parameters in accordance with the number of output gradations. Furthermore, the following problems occur when trying to save the storage capacity of the FIFO memory by using the double error diffusion processing method.
[0024]
When binary, ternary, or quinary processing is performed on an image with 256 gradations, the range of the quantization difference value ΔX generated for each pixel is as follows.
When binarized: -128 to +127
Tri-level: -64 to +63
Five-valued: -32 to +31
[0025]
When δX is diffused to surrounding pixels according to the error diffusion matrix shown in FIG. 1B, the number of bits necessary to express βX stored in the FIFO memory is 8, 7 for binary, ternary, and quinary values, respectively. And 6 bits. If double error diffusion processing is applied in such a case and β′X is obtained by uniformly shifting L bits to the right, the number of effective bits of β′X differs depending on the quantization number. For example, if L = 4, the number of effective bits of β′X is only 6−4 = 2 bits in the case of quinarization.
[0026]
In other words, when using the same FIFO memory in any of binarization, ternarization, and quinarization, when the number of effective bits of β'X is the maximum, that is, according to the number of effective bits of binarization The number of bits of the FIFO memory is determined. If L = 4, 8-4 = 4 bits. Therefore, the two bits of the FIFO memory are not used at the time of quinarying, and the problem that the image quality of the output image is lowered correspondingly is left.
[0027]
[Constitution]
FIG. 2 is a block diagram illustrating a configuration example of an image processing apparatus that performs error diffusion processing according to the present invention. In accordance with the value set in the register of the output gradation number setting unit 211, the gradation number of the output image is binary, Can switch to ternary or quinary.
[0028]
In the figure, reference numeral 1 denotes an image input unit for inputting image data, which is, for example, an apparatus such as an image reader for reading an image and outputting 8-bit digital image data in a raster scanning format. An error diffusion processing unit 2 performs an error diffusion process, which will be described later, on the image data output from the input unit 1. 3 is an image output unit based on the image data output from the error diffusion processing unit 2, such as a monitor that displays an image, a printer that forms an image on recording paper, or an interface that outputs image data to a communication path. is there.
[0029]
The error diffusion processing unit 2 has the following configuration. A quantization unit 202 that performs quantization processing and calculates a quantization difference value, an output gradation number setting unit 211 for setting the number of output gradations, and a FIFO memory 210 that can hold image data for one line ( However, each cell of the FIFO memory is assumed to be 4 bits), adders 201 and 206, multiplier 205, registers 204 and 208, shift registers 203, 207 and 209.
[0030]
[Operation]
FIG. 3 is a flowchart for explaining the error diffusion processing of this embodiment. Assume that the number of pixels for one line of the input image is N, and the error diffusion matrix shown in FIG. 1B is used. 2 is based on the premise that the error diffusion matrix of FIG. 1B is used.If an error diffusion matrix different from FIG. 1B is used, it is necessary to slightly change the configuration of the error diffusion processing unit 2. It does not detract from the essence of the present invention.
[0031]
In step S1, the FIFO memory 210 and all registers and shift registers are cleared to zero. Then, the value of the target pixel X is acquired from the image input unit 1 in step S2 with the upper left pixel of the input image as the target pixel.
[0032]
Next, in step S3, the value of the pixel of interest X acquired from the image input unit 1, the value of the register 204, and the value of the shift register 209 are summed by the adder 201, and the result σX is input to the quantization unit 202. In step S4, the quantization unit 202 performs a quantization process on σX by the following algorithm to obtain a quantization difference value ΔX. The quantization threshold is set by the gradation number setting unit 211 according to the set gradation number.
[0033]
(1) When the number of output gradations is binary: Since the output is 0,1 and the corresponding gray value is 0,255, if σX> 128, output `` 1 '', and δX = σX-255 To do. If σX ≦ 128, “0” is output, and ΔX = σX.
[0034]
(2) When the number of output gradations is ternary: The output is 0, 1, 2 and the corresponding grayscale values are 0, 128, 255, so if σX> 192, output `` 2 '' and δX = σX-255 And If 64 <σX ≦ 192, “1” is output, and ΔX = σX−128. If σX ≦ 64, “0” is output, and ΔX = σX.
[0035]
(3) When the number of output gradations is quinary: The output is 0, 1, 2, 3, 4 and the corresponding gray values are 0, 64, 128, 192, 255, so if σX> 244, output “4” And ΔX = σX−255. If 160 <σX ≦ 244, “3” is output, and ΔX = σX−192. If 96 <σX ≦ 160, “2” is output and ΔX = σX−128. If 32 <σX ≦ 96, “1” is output and ΔX = σX−64. If σX ≦ 32, “0” is output, and ΔX = σX.
[0036]
The quantization result is output to the image output unit 3, and the quantization difference value ΔX is stored in the shift register 203.
[0037]
In step S5, the data read from the FIFO memory 210 is stored in the shift register 209, and the cell of the FIFO memory 210 is advanced by one. In step S6, the shift register 209 is shifted left by L bits. Here, L varies depending on the number of output gradations, and details thereof will be described in step S11.
[0038]
Next, in step S7, ΔX stored in the shift register 203 is shifted right by 2 bits. In step S8, the data stored in the shift register 203 is doubled by the multiplier 205, and the result is input to the adder 206. In step S9, the adder 206 adds the values of the registers 204 and 208 and the output of the adder 206, and stores the result in the shift register 207.
[0039]
Next, in step S 10, the data stored in the shift register 203 is stored in the register 204. In step S11, the shift register 207 is shifted right by L bits, and the result is stored in the FIFO memory 210 and the carry value generated by the processing is stored in the register 208.
[0040]
Here, when the input image data is 8 bits, the number of bits necessary to represent δX is 8 bits when the output gradation number is binary, 7 bits when ternary, and 6 bits when quinary . Therefore, L is determined so that the result of the shift fits in the width of the FIFO memory (4 bits). Specifically, L for binary, ternary, and quinary is 4, 3, and 2, respectively. It is set by the output gradation number setting unit 211.
[0041]
When image data as shown in FIG. 1A is input and the pixel H is the target pixel, the contents of the registers and the shift registers at the end of step S11 are as follows.
Register 204: δH × 1/4
Shift register 207: δG × 1/4 + δH × 2/4
Shift register 209: int {(δ C × 1/4 + δ D × 1/2) / (2 ^ L)} × (2 ^ L)
Register 208: (δG × 1/4 + δH × 2/4)
Carry value generated by L-bit right shift
However, int () represents a truncation operation after the decimal point.
In step S12, it is determined whether or not the target pixel is the right end of the line. If so, the process proceeds to step S13, and if not, the process proceeds to step S14. When the process proceeds to step S13, each register and each shift register are cleared to 0, the target pixel is moved to the left end of the next line, and the process returns to step S2. If the process proceeds to step S14, the target pixel is moved to the next pixel and the process returns to step S2.
[0043]
As described above, in the conventional error diffusion processing, the data stored in the FIFO memory as it is is shifted to the right by L bits and then stored, and the value read from the FIFO memory is shifted to the left by L bits to be used for the subsequent arithmetic processing. By using a series of processes that are used, the number of bits in the FIFO memory cells required for error diffusion processing is reduced, and the number of bits to be shifted is changed according to the number of output gradations. Quantization processing can be performed.
[0044]
That is, the shift amount L of βX is changed according to the quantization number of the error diffusion process. Set L = 4 for binarization, L = 3 for ternarization, L = 2 for quinary, etc. By performing such processing, when the number of output gradations is large (the number of bits of the quantization error is small), the quantization error when quantizing βX to β′X is reduced and more accurate (two Since it is possible to perform an error diffusion process (similar to a conventional error diffusion process that is not a multiple error diffusion process), an improvement in image quality can be expected.
[0045]
In addition, there are two different methods for error diffusion processing in handling data input to the FIFO memory. In the first method, as described above, an error value added from a pixel in the previous line of a line including the pixel is summed in advance and then stored in the FIFO memory. This is called the accumulation method.
[0046]
In the second method, in FIG. 1C, the quantization difference values generated by the quantization processing of the pixels B and C are sequentially stored in the FIFO memory as they are. Then, at the time of quantization processing of the pixel G in the next line, the quantization difference value of the pixels B and C read from the FIFO memory and the quantization difference value of the pixel F are multiplied by the matrix coefficient of the error diffusion matrix. Then, these multiplication results and the value of the pixel G are added to perform the quantization processing of the pixel G, which is called an average error reduction method.
[0047]
The error diffusion method of the present invention can be applied to both the error accumulation method and the average error reduction method. In other words, L bits are shifted right before storing in the FIFO memory, and L bits are shifted left when reading. Similarly, the method of the present invention in which the number of bits L to be shifted is changed according to the number of output gradations can be applied not only to the error accumulation method but also to the average error reduction method.
[0048]
[Other Embodiments]
Note that the present invention can be applied to a system including a plurality of devices (for example, a host computer, an interface device, a reader, a printer, etc.), or a device (for example, a copier, a facsimile device, etc.) including a single device. You may apply to.
[0049]
Another object of the present invention is to supply a storage medium storing software program codes for realizing the functions of the above-described embodiments to a system or apparatus, and the computer (or CPU or MPU) of the system or apparatus stores the storage medium Needless to say, this can also be achieved by reading and executing the program code stored in the. In this case, the program code itself read from the storage medium realizes the functions of the above-described embodiments, and the storage medium storing the program code constitutes the present invention. As a storage medium for supplying the program code, for example, a floppy 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, or the like can be used.
[0050]
Further, by executing the program code read by the computer, not only the functions of the above-described embodiments are realized, but also an OS (operating system) running on the computer based on the instruction of the program code. It goes without saying that a case where the function of the above-described embodiment is realized by performing part or all of the actual processing and the processing is included.
[0051]
Further, after the program code read from the storage medium is written into a memory provided in a function expansion card inserted in the computer or a function expansion unit connected to the computer, the function expansion is performed based on the instruction of the program code. It goes without saying that the CPU or the like provided in the card or the function expansion unit performs part or all of the actual processing and the functions of the above-described embodiments are realized by the processing.
[0052]
When the present invention is applied to the above-described storage medium, the storage medium stores program codes corresponding to the above-described flowcharts. To briefly describe, each module shown in the memory map example of FIG. Is stored in a storage medium. That is, at least the program code of each module of “module name 1”, “module name 2”, and “module name 3” may be stored in the storage medium.
[0053]
【The invention's effect】
As described above, according to the present invention, when performing quantization by error diffusion processing, image processing capable of preventing image quality deterioration of an output image and saving memory capacity of a memory for storing quantization errors. An apparatus and method thereof can be provided.
[Brief description of the drawings]
FIG. 1A is a diagram for explaining an error diffusion method;
FIG. 1B is a diagram for explaining an error diffusion method;
FIG. 1C is a diagram for explaining an error diffusion method;
FIG. 1D is a diagram for explaining an error diffusion method;
FIG. 2 is a block diagram showing a configuration example of an image processing apparatus that performs error diffusion processing according to the present invention;
FIG. 3 is a flowchart for explaining error diffusion processing according to the embodiment;
FIG. 4 is a diagram showing an example of a memory map of a storage medium in which program codes for error diffusion processing according to the present invention are stored.

Claims (8)

An image processing method for performing error diffusion processing on input multilevel image data,
A setting step for setting a threshold and a constant according to the set output gradation number;
A quantization step of quantizing the pixel data with the threshold and determining a quantization error thereof;
A predetermined calculation is performed on a plurality of quantization errors generated by quantization in a certain line, an error integrated value added to data of a predetermined pixel in the next line of the line is calculated, and the error integrated value is divided by the constant. And a first calculation step for storing in the memory,
A second calculation step of multiplying the data read from the memory by the constant and adding to the data of the predetermined pixel,
The image processing method according to claim 1, wherein the quantization step quantizes the pixel data subjected to the operation in the second operation step. An image processing method for performing error diffusion processing on input multilevel image data,
A setting step for setting a threshold and a constant according to the set output gradation number;
A quantization step of quantizing the pixel data with the threshold and determining a quantization error thereof;
A first calculation step of dividing a quantization error generated by quantization by the constant and storing it in a memory;
A second calculation step of multiplying the data read from the memory by the constant, performing a predetermined calculation, and adding to the pixel data;
The image processing method according to claim 1, wherein the quantization step quantizes the pixel data subjected to the operation in the second operation step. 3. The image processing method according to claim 1, wherein the constant is set so that a result of division by the first calculation step is within a predetermined number of bits. An image processing apparatus that performs error diffusion processing on input multi-valued image data,
Setting means for setting a threshold value and a constant according to the set output gradation number;
Quantization means for quantizing pixel data with the threshold and obtaining a quantization error thereof;
A predetermined calculation is performed on a plurality of quantization errors generated by quantization in a certain line, an error integrated value added to data of a predetermined pixel in the next line of the line is calculated, and the error integrated value is divided by the constant. And first calculating means for storing in the storage means,
Second arithmetic means for multiplying the data read from the storage means by the constant and adding to the data of the predetermined pixel,
The image processing apparatus characterized in that the quantization means quantizes pixel data that has been subjected to computation by the second computation means. An image processing apparatus that performs error diffusion processing on input multi-valued image data,
Setting means for setting a threshold value and a constant according to the set output gradation number;
Quantization means for quantizing pixel data with the threshold and obtaining a quantization error thereof;
A first arithmetic unit that divides a quantization error caused by quantization by the constant and stores it in a storage unit;
A second calculation means for multiplying the data read from the storage means by the constant, performing a predetermined calculation, and adding to the pixel data;
The image processing apparatus characterized in that the quantization means quantizes pixel data that has been subjected to computation by the second computation means. 6. The image processing apparatus according to claim 4, wherein the constant is set so that a result of division by the first arithmetic unit is within a predetermined number of bits. A recording medium in which a program code of image processing for performing error diffusion processing on input multi-valued image data is recorded,
A code of a setting step for setting a threshold and a constant according to the set number of output gradations;
A quantization step code for quantizing the pixel data with the threshold and determining the quantization error;
A predetermined operation is performed on a plurality of quantization errors generated by quantization in a line, an error integrated value to be added to data of a predetermined pixel in the next line of the line is calculated, and the error integrated value is divided by the constant. Code of the first calculation step to be stored in the memory,
A code of a second calculation step of multiplying the data read from the memory by the constant and adding to the data of the predetermined pixel,
The recording step is characterized in that the quantization step quantizes the pixel data that has been subjected to the computation in the second computation step. A recording medium in which a program code of image processing for performing error diffusion processing on input multi-valued image data is recorded,
A code of a setting step for setting a threshold and a constant according to the set number of output gradations;
A quantization step code for quantizing the pixel data with the threshold and determining the quantization error;
A code of a first calculation step for dividing a quantization error caused by quantization by the constant and storing it in a memory;
A second operation step code for multiplying the data read from the memory by the constant, applying a predetermined operation to the pixel data, and
The recording step is characterized in that the quantization step quantizes the pixel data that has been subjected to the computation in the second computation step.

Images