JP2883146B2 - Binarization method of gradation image - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing method for binarizing a multi-value gradation image in a scanner, a printer, a facsimile, a digital copy, or the like.

(Prior Art) Conventionally, in image processing, data processing for binarizing the brightness of an image into 0 and 1 has been performed.

FIG. 4 is a block diagram schematically illustrating an example of the image processing apparatus. In the figure, 41 is an image input device, 42 is a color conversion correction device, 43 is a sharpness correction device, 44 is an image binarization device, and 45 is an output device.

The input gradation image data from the image input device 41 is subjected to complementary color conversion and color correction by a color conversion correction device 42, and is subjected to necessary sharpness correction by a sharpness correction device 43.
The image is converted by the image binarizing device 44 into binarized data corresponding to the brightness. The converted data is printed by the output device 45, for example, a color printer.

In this way, in image processing, binarization of pixels in an image having a gray scale assumes that each pixel has a desired lightness between 0 and 1, and scans the image line by line, Is binarized into values of 1 and 0 corresponding to light and dark. In this case, pixels whose desired brightness is 0.5 or less
Assuming that the lightness is 0 or 1, the pixel whose desired lightness is 0 or 1 is binarized to 0 or 1 exactly.
Pixels between 0 and 1 are truncated or rounded up and converted to a brightness of 0 or 1, so the intermediate brightness is too brighter than the actual brightness, or
It is binarized as a value that is too dark. Therefore, pixels of intermediate brightness where the desired brightness is not 0 or 1 will cause errors in the image.

Floyd's algorithm (R.Fioyd & L.Stei
nberg: "An Adaptive Algorithm for Spatial greyscal
e ", SID Symp., Digest of Papers, p.36 (1975)), the error diffusion method binarizes and distributes the error generated at that pixel to some surrounding pixels that have not been scanned yet. To correct the brightness.

FIG. 5 is an explanatory diagram of one example. In the figure, each square is one pixel. The cross-hatched pixels 52 are pixels under processing, and the hatched pixels 53 are binarized pixels. The other non-hatched pixels 54 are unprocessed pixels. The direction of pixel processing in each scanning line, which is the direction of main scanning, is
As shown by an arrow 51, the scanning is performed sequentially from left to right on all the scanning lines. An error, which is a difference from the binarized data in the pixel 52 being processed, is distributed to four pixels according to the weighting factors A to D, for example, as shown in FIG. As shown in FIG. 7A, the distribution destination is 7/16 of the error for the pixel A to be processed next, and 1/16, for the pixels B to D located before and after the next scanning line, respectively. 5/16 and 3/16 are distributed. Note that the position shown as K indicates a texture shift in a direction called a K point.

FIG. 3 is a schematic configuration diagram of an example that embodies the above-described error diffusion method. In the figure, 1 is an image input, 2 is an input buffer memory, 3 is an adder, 4 is a threshold setting circuit, 5 is a comparator, 6 is a calculator, 7 is a weighting circuit, 8 is an error buffer memory, and 9 is an output buffer. The memory 10 is a binarized output.

Image input 1 is stored in input buffer memory 2 and
It is read out pixel by pixel. After passing through the adder 3, the comparator 5 compares the threshold value with a constant threshold value from the threshold value setting circuit 4, for example, 0.5, which is a lightness equivalent value, and binarizes it to 0 or 1 according to the magnitude thereof. Stored in the memory 9. The difference between the data input in the comparator 5 and the threshold value is calculated in the calculator 6, and based on the obtained error, the weighting circuit 7 calculates the weight coefficient shown in FIG. 7 is stored. The error data stored in the error buffer memory 7 is stored as a cumulative value for unprocessed pixels, and is distributed according to the pixels introduced into the adder 3. Therefore, the binarized output 10 stores binarized data based on a value that takes into account the distributed error data.

The problem with this algorithm is that texture shifts in the direction called the K point and linear noise due to a phenomenon called the history effect occur, and due to this noise, the best binarization performance is achieved. However, there were few examples of application.

Further, the error buffer 8 for several lines must be provided, and the amount of calculation is large, so that the apparatus is complicated and expensive.

In order to simplify the apparatus, as shown in FIG. 6, a method of diffusing an error only in a one-dimensional direction can be considered. According to this method, a memory capacity of data for several pixels is sufficient, so that the apparatus is simplified. However, as shown in FIG.
However, there is a problem in that the pattern 71 is likely to appear with regularity in the direction perpendicular to the processing direction and has a stripe pattern.

(Problems to be Solved by the Invention) The present invention has been made in order to solve the above-described problems, and enables high-speed processing and simplification of a circuit without deteriorating an image, and is fast and inexpensive. It is an object of the present invention to provide a binarization method that can be configured as follows.

(Means for Solving the Problems) The present invention binarizes pixel data of image information having a gradation, and distributes an error component generated with the binarization of the pixel data to a predetermined distribution ratio, or In a method of binarizing a grayscale image, the pixel data is binarized by diffusing and distributing to peripheral unprocessed pixel data by a random distribution ratio, and adding up the sum of the distributed errors. The threshold value in the binarization is changed periodically or randomly for each line, and the error to be distributed is initialized for each predetermined number of pixels or with a random number of pixels. Things.

(Operation) The pixel data of the image information having a gradation is binarized, and an error component generated due to the binarization of the pixel data is processed in a peripheral unprocessed state by a preset distribution ratio or a random distribution ratio. In a binarization method of a gradation image in which pixel data is diffused and distributed, and a cumulative total of the distributed errors is added to perform binarization of the pixel data, if a threshold value in the binarization is changed, Even at the level of image data,
The position of the binarized dot changes. Therefore, by changing the threshold periodically or randomly for each line, the positions where dots appear can be dispersed, and a high-quality image can be obtained even if a simple error distribution coefficient and distribution destination are used. Also, by initializing the distributed error for each predetermined number of pixels or with a random number of pixels, the effect of threshold change due to accumulation of errors can be maintained in an effective initial state.

(Embodiment) FIG. 1 is a schematic configuration diagram that embodies a method of changing a threshold value in an embodiment of the above-described error diffusion method of the present invention. Parts similar to those in FIG. 3 are denoted by the same reference numerals, and description thereof is omitted. 3 in that a delay circuit 11 is used without providing the error buffer 8 and that a threshold memory 12 and a line synchronization control circuit 13 for controlling the threshold setting circuit 4 are provided. .

The reason for using the delay circuit 11 is that the error is distributed to the same direction as the main scanning direction, that is, on the main scanning line, as shown in FIG. 6, and the distribution destination is only the next pixel. That is, in FIG. 6, A is set to 1. As a result, the error data can be delayed and the weighting factor can be added by the weighting circuit 7, so that there is no need to use an error buffer memory, and the circuit and the operation become extremely simple.

A threshold memory 12 for controlling the threshold setting circuit 4 and a line synchronization control circuit 13 will be described together with a binarization method. The input image data once stored in the input buffer 2 is added to the error data of the processed pixels by the adder 3, and is compared with a threshold. The threshold is read from the threshold memory 12 for each line by the line synchronization control circuit 13 and set. The binary data binarized by the comparator 5 is sent to the output buffer 9 and becomes a binarized output. At this time, a signal obtained by subtracting the binarized output from the comparator 5 from the signal before binarization becomes an error in this pixel.
The error calculated by the arithmetic unit 6 is 1 by the delay circuit 11.
After being delayed by the number of pixels, it is multiplied by a predetermined weighting coefficient by the weighting circuit 7 and input to the adder 2-3. At this time, since the error is delayed by one pixel, the adder 3 simply adds the next pixel.

The operation of the binarization device shown in FIG. 1 will be described in more detail with reference to FIGS. 9 and 10.

FIG. 9 shows pixel data, error data, and binarized print data when halftone pixels with a density of 1/4 are input to a line where the threshold is set to 1/2. FIG. 4 is a diagram schematically representing In pixels, the input value 1/4 is not printed because it does not exceed the threshold value 1/2. As for the error at the pixel, the output is 0 (not printed) for 1/4 of the input value, so 1 / 4-0 = 1/4 is passed to the pixel as it is. At the pixel, 1/4 of the input is added to 1/4 of the previous error, giving 1/4 + 1/4 = 1/2, which is compared to 1/2 the threshold (if equal, the excess ), The output becomes 1 and it is printed. At this time, the error was printed for 1/2 of the input, so the output was 1, and 1 / 2−1 = −1 / 2. From the pixel, 1/4 of the input is added to the previous error -1/2, and thereafter, the error is propagated as -1 / 4,0,1 / 4, and at the pixel, Is printed. ..,..., One pixel is printed every four pixels, and a density of 1/4 is expressed.

Here, if the same threshold value is processed for the next line, the printing position becomes the same, and it can be easily understood that the conventional method having vertical stripes shown in FIG. 7 is used.
In this embodiment, the threshold level is changed in the next line. However, it is not always necessary to change for each line.

FIG. 10 schematically shows pixel data, error data, and print data when a halftone having a density of 1/4 is input to the next line where the threshold value is set to 1, as in FIG. FIG. 1/4 error from pixel to pixel
Accumulates one by one and the input is 1
/ 4, 3/4 + 1/4 = 1, and exceeds threshold 1, so it is printed. Since the error generated at this time is output 1 with respect to input 1, 1-1 = 0, and the error passed to the pixel is 0. Therefore, after the pixel, the operation from the pixel to the pixel is repeated, and one is printed for every four pixels, such as 4, 8, 12, 16, ..., and a density of 1/4 is expressed. .

By repeating the threshold value of each line as 1/2, 1, 1/2, 1, in this manner, the halftone print output becomes as shown in FIG. 8, and the vertical stripes shown in FIG. Binarization can be performed without any problem

FIG. 11 shows an example of a pattern when the threshold value is changed for each line. (A) shows a case where two lines are repeated as a cycle, (B) shows a case where four lines are repeated, (C) shows a case where eight lines are repeated, and (D) shows 16 lines. Is a cycle. The period is preferably large to some extent in order to uniformly print low-density halftones, and generally, 4 to 8 lines or more is desirable.

The threshold value change for each line may be random instead of periodic. By making them random, image noise having a periodically fixed spatial frequency component can be made more inconspicuous.

FIG. 2 is a schematic configuration diagram that embodies an embodiment of the error diffusion method of the present invention. Parts similar to those in FIG. 1 are denoted by the same reference numerals, and description thereof is omitted.

In this embodiment, an initialization circuit 14 controlled by an initialization control circuit 15 is provided on the output side of the weighting circuit 7 in FIG. In the apparatus shown in FIG. 1, there is no problem at the beginning of the process, that is, at the beginning of the line. However, as the process progresses, as the error accumulates, the effect of changing the threshold value becomes weaker, and gradually. Vertical stripes may occur. Therefore, in this embodiment, for example,
The accumulated error is cleared to 0 every several pixels. That is, initialization is performed to return to an initial state where no error is accumulated. Thereby, the effect of the threshold value change can be maintained in an effective initial state.

The initialization control circuit 15 controls how many pixels are to be initialized. This initialization cycle is about the number of input gradations, but the threshold value of each line is periodically changed. Is set as a guideline, and it is desirable to set the period more than that.

According to the present invention, the error distribution destination can be simplified and can be set only in the line direction, so that the binarizing device can be constituted by a delay circuit instead of an expensive error buffer.

In the above-described embodiment, the error is distributed to the same line. However, the error may be distributed to different lines as shown in FIG. Even in this case, a sufficient effect can be obtained even if the allocation destination and the allocation coefficient are simple compared to the conventional method.

(Effects of the Invention) As is clear from the above description, according to the present invention, it is possible to increase the processing speed and simplify the circuit. There is an effect that a binarization method can be provided.

[Brief description of the drawings]

FIG. 1 is a schematic configuration diagram that embodies a method of changing a threshold value in one embodiment of the error diffusion method of the present invention, and FIG.
Schematic configuration diagram embodying an embodiment of the error diffusion method of the present invention,
FIG. 3 is a schematic configuration diagram of a conventional error diffusion method, and FIG.
5 is a block diagram for explaining an outline of the image processing apparatus, and FIG.
FIGS. 6 and 7 are explanatory diagrams of weighting factors, FIGS. 7 and 8 are explanatory diagrams of an example of a print dot, and FIGS. 9 and 10 are explanatory diagrams of a binarization operation when a threshold value is changed. FIG. 11 is an explanatory diagram of a pattern when the threshold value is changed for each line. 1 image input, 2 input buffer memory, 3 adder, 4 threshold setting circuit, 5 comparator, 6 arithmetic unit, 7 weighting circuit, 8 error buffer memory ,
9 ... output buffer memory, 10 ... binarized output, 11 ...
Delay circuit 12, threshold memory, 13 line synchronization control circuit, 14 initialization circuit, 15 initialization control circuit.

Claims (1)

(57) [Claims] 1. A method according to claim 1, wherein pixel data of image information having a gradation is binarized, and an error component generated in accordance with the binarization of the pixel data is divided into peripheral components by a preset distribution ratio or a random distribution ratio. In a binarization method of a grayscale image in which pixel data is diffused and distributed to unprocessed pixel data, and a total of the distributed errors is added to perform binarization of the pixel data, a threshold in the binarization is set to a line A binarization method of a gradation image, wherein the error to be distributed is changed periodically or randomly, and an error to be distributed is initialized every predetermined number of pixels or with a random number of pixels. .