WO2023206081A1

WO2023206081A1 - Content-adaptive random spatial dithering

Info

Publication number: WO2023206081A1
Application number: PCT/CN2022/089321
Authority: WO
Inventors: Teppei Isobe
Original assignee: Huawei Technologies Co.,Ltd.
Priority date: 2022-04-26
Filing date: 2022-04-26
Publication date: 2023-11-02

Abstract

An embodiment of the present invention provides a device for dithering using a Bayer matrix. The device includes: a characteristic detection unit configured to detect a characteristic of an original image; a control parameter unit configured to determine a control value based on the characteristic of the original image; a first adder configured to add the control value to respective elements of the Bayer matrix to obtain modified elements of the Bayer matrix; a second adder configured to repeatedly add a modified element of the Bayer matrix to a corresponding pixel value of the image to obtain a modified pixel value, and a shifter configured to shift the modified pixel value to obtain an image expressed with decreased bit depth. The characteristic of the original image may be detected by using image frequency histogram. The respective elements of the Bayer matrix may be respective original elements of the Bayer matrix, or respective elements obtained by multiplying the respective original elements of the Bayer matrix by a factor that is determined based on the characteristic of the original image. The control parameter unit may be further configured to determine a control gain based on the characteristic of the original image, and the device further comprises a first multiplier configured to multiply random noise by the control gain to obtain modified random noise; and a third adder configured to, before repeatedly adding the modified element of the Bayer matrix to the corresponding pixel value of the image, update the respective modified elements of the Bayer matrix by adding the modified random noise to the respective modified elements of the Bayer matrix. The embodiment of the present invention achieves improving quality of the image.

Description

CONTENT-ADAPTIVE RANDOM SPATIAL DITHERING

TECHNICAL FIELD

The present application relates to a device and a method for dithering for images.

BACKGROUND

Display devices such as an electrophoretic display (EPD) can’t show a high bit depth image. Most of these devices use 4 bits for each pixel. In these devices, in order to achieve high refresh rate, bit depth will be decreased to 2 bits or 1 bit, and image quality is not enough. For example, if an 8 bit image is converted into a 2 bit image, grayscale of large areas becomes flat.

In order to express grayscale accurately, dithering technology is used. There are basically two types: one is for spatial domain, and another is for time domain. Because the time domain technology cannot be used for EPD devices, the following will focus on the spatial domain technology. There are mainly two types of dithering technology for spatial domain. One is an error diffusion type, and another is a Bayer matrix type. In many cases, “Floyd-Steinberg” dithering is used in the error diffusion type.

Regarding the error diffusion type, FIG. 1 shows a basic concept of error diffusion, namely, how to spread a quantization error in image data. A quantization error of a pixel is added onto its neighboring pixels. The pixel indicated with “*” indicates the pixel currently being scanned. The image is scanned from left to right and top to bottom, pixel values are quantized one by one, and the quantization error is transferred to the neighboring pixels, namely, the right pixel, the lower left pixel, the lower pixel, and the lower right pixel in the ratio shown in Fig. 1.

Regarding the Bayer matrix type, FIG. 2 shows an example of a Bayer matrix. This matrix is repeatedly placed on the image data to determine whether the quantized pixel value is incremented or not. Namely, assuming that a pixel value is quantized to an integer between 0 to 255, the value of the corresponding position in the Bayer matrix is divided by 16, and if the quantization error is higher than the divided value, the quantized pixel value is incremented by one.

In the error diffusion type, worm noise occurs in the image as shown in Fig. 3 (a) , and it needs a line memory to store data for the other pixel. In the Bayer matrix type, fixed pattern noise occurs in the image as shown in Fig. 3 (b) . In this way, prior techniques cause much noise in the images.

SUMMARY

A device for dithering is provided to achieve improving quality of the image, in particular, flat areas of the image.

According to a first aspect, a device for dithering using a Bayer matrix is provided, where the device includes: a characteristic detection unit configured to detect a characteristic of an original image; a control parameter unit configured to determine a control value based on the characteristic of the original image; a first adder configured to add the control value to respective elements of the Bayer matrix to obtain modified elements of the Bayer matrix; a second adder configured to repeatedly add a modified element of the Bayer matrix to a corresponding pixel value of the image to obtain a modified pixel value; and a shifter configured to shift the modified pixel value to obtain an image expressed with decreased bit depth.

The respective elements of the Bayer matrix are adjusted based on the characteristic of the original image, and thus quality of the image expressed with decreased bit depth can be improved.

In a possible implementation of the first aspect, the characteristic of the original image is detected by using image frequency histogram. The characteristic of the original image can be appropriately judged by classifying a tendency of frequency components of the image.

In a possible implementation of the first aspect, the respective elements of the Bayer matrix are respective original elements of the Bayer matrix, or respective elements obtained by multiplying the respective original elements of the Bayer matrix by a factor that is determined based on the characteristic of the original image. The respective elements of the Bayer matrix are adjusted based on the factor that depends on the characteristic of the original image, and thus quality of the image expressed with decreased bit depth can be improved.

In a possible implementation of the first aspect, the control parameter unit is further configured to determine a control gain based on the characteristic of the original image, and the device further includes a first multiplier configured to multiply random noise by the control gain to obtain modified random noise; and a third adder configured to, before repeatedly adding the modified element of the Bayer matrix to the corresponding pixel value of the image, update the respective modified elements of the Bayer matrix by adding the modified random noise to the respective modified elements of the Bayer matrix. The random noise added to the respective elements of the Bayer matrix is adjusted based on the characteristic of the original image, and thus quality of the image expressed with decreased bit depth can be improved.

In a possible implementation of the first aspect, the control parameter unit is further configured to determine a pre gain based on the characteristic of the original image, and the device further includes a second multiplier configured to, before repeatedly adding the modified element of the Bayer matrix to the corresponding pixel value of the image, update respective pixel values by multiplying the respective pixel values of the image by the pre gain. The respective pixel values of the image are adjusted based on the characteristic of the original image, and thus quality of the image expressed with decreased bit depth can be improved.

In a possible implementation of the first aspect, the amount of noises, which are caused by dithering, in the image expressed with decreased bit depth differs depending on the size of flat area when using original images with different sizes of flat areas.

According to a second aspect, a device for dithering using error diffusion is provided, where the device includes: a characteristic detection unit configured to detect a characteristic of an original image; a control parameter unit configured to determine a control value based on the characteristic of the original image; a first adder configured to add the control value to respective coefficients for the error diffusion to obtain modified coefficients for the error diffusion; a second adder configured to add quantization errors, which are calculated based on the modified coefficients for the error diffusion, from neighboring pixels to each pixel value of the image to obtain a modified pixel value; and a shifter configured to shift the modified pixel value to obtain an image expressed with decreased bit depth.

The respective coefficients for the error diffusion are adjusted based on the characteristic of the original image, and thus quality of the image expressed with decreased bit depth can be improved.

In a possible implementation of the second aspect, the characteristic of the original image is detected by using image frequency histogram. The characteristic of the original image can be appropriately judged by classifying a tendency of frequency components of the image.

In a possible implementation of the second aspect, the respective coefficients for the error diffusion are respective original coefficients for the error diffusion, or respective coefficients obtained by multiplying the respective original coefficients for the error diffusion by a factor that is determined based on the characteristic of the original image. The respective coefficients for the error diffusion are adjusted based on the characteristic of the original image, and thus quality of the image expressed with decreased bit depth can be improved.

In a possible implementation of the second aspect, the control parameter unit is further configured to determine a control gain based on the characteristic of the original image, and the device further includes a first multiplier configured to multiply random noise by the control gain to obtain modified random noise; and a third adder configured to, before adding quantization errors, which are calculated based on the modified coefficients for the error diffusion, from neighboring pixels to each pixel value of the image, update the respective modified coefficients for the error diffusion by adding the modified random noise to the respective modified coefficients for the error diffusion. The random noise added to the respective coefficients for the error diffusion is adjusted based on the characteristic of the original image, and thus quality of the image expressed with decreased bit depth can be improved.

In a possible implementation of the second aspect, the control parameter unit is further configured to determine a pre gain based on the characteristic of the original image, and the device further includes a second multiplier configured to, before adding quantization errors, which are calculated based on the modified coefficients for the error diffusion, from neighboring pixels to each pixel value of the image, update respective pixel values by multiplying the respective pixel values of the image by the pre gain. The respective pixel values of the image are adjusted based on the characteristic of the original image, and thus quality of the image expressed with decreased bit depth can be improved.

In a possible implementation of the second aspect, the amount of noises, which are caused by dithering, in the image expressed with decreased bit depth differs depending on the size of flat area when using original images with different sizes of flat areas.

According to a third aspect, a method for dithering using a Bayer matrix is provided, where the method includes: detecting a characteristic of an original image; determining a control value based on the characteristic of the original image; adding the control value to respective elements of the Bayer matrix to obtain modified elements of the Bayer matrix; repeatedly adding a modified element of the Bayer matrix to a corresponding pixel value of the image to obtain a modified pixel value; and shifting the modified pixel value to obtain an image expressed with decreased bit depth.

In a possible implementation of the third aspect, the characteristic of the original image is detected by using image frequency histogram. The characteristic of the original image can be appropriately judged by classifying a tendency of frequency components of the image.

In a possible implementation of the third aspect, the respective elements of the Bayer matrix are respective original elements of the Bayer matrix, or respective elements obtained by multiplying the respective original elements of the Bayer matrix by a factor that is determined based on the characteristic of the original image. The respective elements of the Bayer matrix are adjusted based on the factor that depends on the characteristic of the original image, and thus quality of the image expressed with decreased bit depth can be improved.

In a possible implementation of the third aspect, the method further includes: determining a control gain based on the characteristic of the original image; multiplying random noise by the control gain to obtain modified random noise; and before repeatedly adding the modified element of the Bayer matrix to the corresponding pixel value of the image, updating the respective modified elements of the Bayer matrix by adding the modified random noise to the respective modified elements of Bayer matrix. The random noise added to the respective elements of the Bayer matrix is adjusted based on the characteristic of the original image, and thus quality of the image expressed with decreased bit depth can be improved.

In a possible implementation of the third aspect, the method further includes: determining a pre gain based on the characteristic of the original image; and before repeatedly adding the modified element of the Bayer matrix to the corresponding pixel value of the image, updating respective pixel values by multiplying the respective pixel values of the image by the pre gain. The respective pixel values of the image are adjusted based on the characteristic of the original image, and thus quality of the image expressed with decreased bit depth can be improved.

In a possible implementation of the third aspect, the amount of noises, which are caused by dithering, in the image expressed with decreased bit depth differs depending on the size of flat area when using original images with different sizes of flat areas.

According to a fourth aspect, a method for dithering using error diffusion is provided, where the method includes: detecting a characteristic of an original image; determining a control value based on the characteristic of the original image; adding the control value to respective coefficients for the error diffusion to obtain modified coefficients for the error diffusion; adding quantization errors, which are calculated based on the modified coefficients for the error diffusion, from neighboring pixels to each pixel value of the image to obtain a modified pixel value; and shifting the modified pixel value to obtain an image expressed with decreased bit depth.

In a possible implementation of the fourth aspect, the characteristic of the original image is detected by using image frequency histogram. The characteristic of the original image can be appropriately judged by classifying a tendency of frequency components of the image.

In a possible implementation of the fourth aspect, the respective coefficients for the error diffusion are respective original coefficients for the error diffusion, or respective coefficients obtained by multiplying the respective original coefficients for the error diffusion by a factor that is determined based on the characteristic of the original image. The respective coefficients for the error diffusion are adjusted based on the characteristic of the original image, and thus quality of the image expressed with decreased bit depth can be improved.

In a possible implementation of the fourth aspect, the method further includes: determining a control gain based on the characteristic of the original image; multiplying random noise by the control gain to obtain modified random noise; and before adding quantization errors, which are calculated based on the modified coefficients for the error diffusion, from neighboring pixels to each pixel value of the image, updating the respective modified coefficients for the error diffusion by adding the modified random noise to the respective modified coefficients for the error diffusion. The random noise added to the respective coefficients for the error diffusion is adjusted based on the characteristic of the original image, and thus quality of the image expressed with decreased bit depth can be improved.

In a possible implementation of the fourth aspect, the method further includes: determining a pre gain based on the characteristic of the original image; and before adding quantization errors, which are calculated based on the modified coefficients for the error diffusion, from neighboring pixels to each pixel value of the image, updating respective pixel values by multiplying the respective pixel values of the image by the pre gain. The respective pixel values of the image are adjusted based on the characteristic of the original image, and thus quality of the image expressed with decreased bit depth can be improved.

In a possible implementation of the fourth aspect, the amount of noises, which are caused by dithering, in the image expressed with decreased bit depth differs depending on the size of flat area when using original images with different sizes of flat areas.

According to a fifth aspect, a device for dithering using a Bayer matrix is provided, where the device includes a processor and a memory, wherein the processor is coupled to the memory, the memory is configured to store a computer program, and the processor is configured to execute the computer program in the memory, to perform the method according to the third aspect or any one of the possible implementations of the third aspect.

According to a sixth aspect, a device for dithering using a Bayer matrix, comprising a processor and a memory, wherein the processor is coupled to the memory, the memory is configured to store a computer program, and the processor is configured to execute the computer program in the memory, to perform the method according to the fourth aspect or any one of the possible implementations of the fourth aspect.

BRIEF DESCRIPTION OF DRAWINGS

To describe the technical solutions in the embodiments of the present invention or in the prior art more clearly, the following briefly introduces the accompanying drawings required for describing the embodiments or the prior art. The accompanying drawings in the following description merely show some embodiments of the present invention, and a person of ordinary skill in the art may still derive other drawings from these accompanying drawings without creative effort.

FIG. 1 shows a basic concept of error diffusion;

FIG. 2 shows an example of Bayer matrix;

FIG. 3 (a) shows an example of worm noise caused by error diffusion dithering;

FIG. 3 (b) shows an example of fixed pattern noise caused by Bayer matrix dithering;

FIG. 4 shows an exemplary block diagram for 4-bit dithering according to Embodiment 1 of the present invention;

FIG. 5 shows how to determine a pre gain, a control gain, and a control value based on image types;

FIG. 6 shows an example of the pre gain when output grayscale is reduced to 2 bits.;

FIG. 7 shows an exemplary block diagram for dithering according to Embodiment 2 of the present invention;

FIG. 8 shows an exemplary block diagram of a panel control unit 30;

FIG. 9 shows an exemplary block diagram of a device for dithering 40;

FIG. 10 shows an example of a Bayer matrix;

FIG. 11 (a) shows an example image processed by Embodiment 1 of the present invention; and

FIG. 11 (b) shows an example image processed by Embodiment 1 of the present invention.

DESCRIPTION OF EMBODIMENTS

The following clearly and completely describes the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. The described embodiments are only some but not all of the embodiments of the present invention. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protected scope of the present invention.

The embodiments of the present invention add random elements to the image data for dithering. In order to reduce the noise on a flat image area, random values and dithering parameters are adaptively controlled depending on the content. In the embodiments of the present application, random noise is combined with the Bayer matrix dithering to diffuse the pattern noise. A pre gain, a control gain, and a control value are used to adjust the RGB data, random noise, and Bayer matrix based on the image characteristics.

Fig. 4 shows an exemplary block diagram for 4-bit dithering according to Embodiment 1 of the present invention. In Fig. 4, a device for dithering 10 receives N-bit RGB data, and outputs N-4 bit RGB data. Although the number of bits is decreased, the quality of the flat image area on the display can be improved.

A characteristic detection unit 12 receives the N-bit RGB data, and detects the characteristics of the RGB data. For example, as shown in Fig. 5, there are three image types: (1) picture, (2) animation or game, and (3) Web page or article. Among these image types, spatial frequency is higher in the order of (1) , (2) , and (3) . For example, the image type can be judged by image frequency histogram, etc. The image type may be classified based on APL (Average Picture Level) . The characteristic of the image can be appropriately judged by classifying a tendency of frequency components of the image. The image type may be automatically detected, or may be set by a user as an operation mode of an application program.

Based on the detected image type, a control parameter unit 13 determines a pre gain, a control gain, and a control value, outputs the pre gain to a multiplier 11, outputs the control gain to a multiplier 14, and outputs the control value to an adder 15. The multiplier 11 multiplies the RGB data by the pre gain. The multiplier 14 multiplies random noise values by the control gain.

The control value is converted into a minus value and added to respective elements of a Bayer matrix 101 by the adder 15, in other words, the control value is subtracted from the respective elements of the Bayer matrix 101. Before the operation of the adder 15, the respective elements of the Bayer matrix 101 may be multiplied by a factor, for example, 0.75, 0.5, etc. based on the detected image type. The elements of the Bayer matrix may have plus or minus values.

The adder 16 merges the random noise values and the respective elements of the Bayer matrix, in other words, updates the respective elements of the Bayer matrix by adding the output value of the multiplier 14 and the output value of the adder 15 to obtain modified elements of the Bayer matrix. By using the modified elements of the Bayer matrix, 4-bit dithering is performed, namely, the adder 17 adds the output value of the multiplier 11 and the output value of the adder 16, in other words, each modified element of the Bayer matrix is repeatedly added to each pixel value of the image, and then a shifter (shown as “N-4 bit” in Fig. 4) shifts the output value of the adder 17 to the right by 4 bits, so that the output value of the adder 17 is divided by 16 to output N-4 bit RGB data. This constitution of the device for dithering 10 does not need a line memory because data for the other pixel is not used to obtain the output RGB data.

Random noise 200 can be generated by using various algorithms for generating pseudo random numbers.

When displaying moving images with a few bits, for example, 1 or 2 bits per pixel, there may be cases where random noise seems to be flicker. In this case, a random number generator used for generating pseudo random numbers may be reset with the same seed value for each frame.

Finally, a 32-bit random value is obtained by using the above algorithm, and the random noise 200 is generated by repeatedly calculating this random value. Instead of the random noise 200, random noise which has a predetermined percentage of the random noise 200 may be use, for example, random noise 201 which has values of 25 %of the random noise 200, or random noise 202 which has values of 10 %of the random noise 200 may be used.

Fig. 5 shows how to determine the pre gain, the control gain, the control value, and the random noise.

For the image type “ (1) picture” which generally includes high frequency images, the Bayer matrix uses minus values, random element is enhanced, and the dithering level is high. For example, the pre gain is set to 0.7, the control gain is set to 0.02, the control value is set to 4, the random noise 201 (25%of the random noise 200) is selected. The elements of the Bayer matrix 101, which is the base of various versions such as Bayer matrix 102 to 104, etc., ranges from 0 to +15 inclusive. The Bayer matrix 102 is obtained by subtracting 4, which is the control value, from the respective elements of the Bayer matrix 101, namely, the elements of the Bayer matrix 102 range from -4 to +11 inclusive.

For the image type “ (2) animation or game” which generally includes relatively simple images compared to the image type “ (1) picture” , the random element is reduced, and the dithering level is middle. This setting reduces noise on flat areas. For example, the pre gain is set to 0.6, the control gain is set to 0.01, the control value is set to 1, the random noise 202 (10%of the random noise 200) is selected. The Bayer matrix 103 is obtained by multiplying the respective elements of the Bayer matrix 101 by 0.75 and subtracting 1 from the multiplied value, and the elements of the Bayer matrix 103 range from -1 to +10.5 inclusive.

For the image type “ (3) Web or article” which generally includes very simple images compared to the image type “ (1) picture” , the values of the elements of the Bayer matrix are reduced to half, the random element is reduced, and the dithering level is low. This setting reduces noise on flat areas. For example, the pre gain is set to 0.6, the control gain is set to 0.01, the control value is set to 0, the random noise 202 (10%of the random noise 200) is selected, the Bayer matrix 104 is obtained by multiplying the respective elements of the Bayer matrix 101 by 0.5, and the elements of the Bayer matrix 104 range from 0 to +7.5 inclusive.

Fig. 6 shows an example of the pre gain when output grayscale is reduced to 2 bits. Even if input grayscale ranges from 0.75 to 1.0, output grayscale ranges from 0 to 3, and thus 0.75 to 1.0 cannot be expressed. The range of the input grayscale is reduced to 0 to 0.75 by multiplying the input grayscale by 0.75 before the dithering process.

Fig. 7 shows an exemplary block diagram for the dithering according to Embodiment 2 of the present invention. This embodiment combines the random noise and the error diffusion dithering. The way of dithering can be changed to any kinds of methods.

The operations of a multiplier 21, a characteristic detection unit 22, a control parameter unit 23, and a multiplier 24 are the same as the operations of a multiplier 11, a characteristic detection unit 12, a control parameter unit 13, and a multiplier 14 in Fig. 4, respectively. The processing of coefficients for error diffusion is as follows: An adder 25 converts a control value received from a control parameter unit 23 into a minus value, and adds it to respective coefficients for error diffusion, in other words, the control value is subtracted from the respective coefficients for error diffusion. Before the operation of the adder 25, the respective coefficients for error diffusion may be multiplied by a certain factor that is determined based on the detected image type. The coefficients for error diffusion may have plus or minus values ranging from -16 to +15 inclusive, and these values correspond to numerators of fractional values shown in the right side of Fig. 7.

The adder 26 merges the random noise values and the respective coefficients for error diffusion by adding the output value of the multiplier 24 and the output value of the adder 25 to obtain modified coefficients for error diffusion. By using the modified coefficients for error diffusion, 4-bit dithering is performed, namely, a quantization error when quantizing the current pixel value from N bits to N-4 bits is spread to neighboring pixels corresponding to respective coefficients, namely, the right pixel, the lower left pixel, the lower pixel, and the lower right pixel, the values to be spread are stored in a line memory (not shown) , and the adder 27 adds the output value of the multiplier 21 (the current pixel value) and the stored values for the current pixel (quantization errors from neighboring pixels) , and then a shifter (shown as “N-4 bit” in Fig. 7) shifts the output value of the adder 27 to the right by 4 bits, so that division by 16 is performed for the output of an adder 27.

Fig. 8 shows an exemplary block diagram of a panel control unit 30. This constitution can be applicable to any kinds of display devices. Video data is input into an image signal processing (ISP) unit 31 and a characteristic detection unit 32. A panel correction unit 33 performs various kinds of correction processing for the image signal to be displayed, for example, correction processing related to a gamma value, luminance unevenness, luminance degradation, local dimming, etc. A dithering unit 34 performs dithering processing as explained above with reference to Fig. 4 or Fig. 7. The characteristics of image data are obtained from the characteristic detection unit 32, and contents information (corresponding to “image type” in Fig. 5) is obtained from outside of the panel control unit 30. The dithering unit 34 outputs RGB data to a source driver controller 35 and a gate driver controller 36. The source driver controller 35 and the gate driver controller 36 control a source driver 37 and a gate driver 38 to drive a panel 39. It is better that the dithering unit 34 is placed before the source driver controller. However, the position of the dithering unit 34 is not limited to this embodiment.

Fig. 9 shows an exemplary block diagram of a device for dithering 40. The device for dithering 40 includes a processor 41 and a memory 42, the processor 41 is coupled to the memory 42 via a bus 43, the memory 42 is configured to store a computer program, and the processor 41 is configured to execute the computer program in the memory 42, to perform processes described above with reference to Fig. 4 to Fig. 7.

The following describes three examples for comparison between two cases with different dithering for images compressed from 8 bits to 2 bits. The first example is a comparison between prior art technology and Embodiment 1. The second and the third examples are comparisons between different parameters of the Embodiment 1, and show improvement and effectiveness for images by content adaptive parameter control according to the embodiments of the present invention.

For the first example, the image to be processed is a photograph of a building (not shown) , namely, a natural image. In an example image processed by error diffusion (prior art technology) , dark area cannot be expressed, and worm noise occurs in the image. In an example image processed by Embodiment 1 of the present invention, the pre gain is set to 0.7, and the random noise 201 (25%of the random noise 200) is selected. Bayer matrix 105 shown in Fig. 10 is obtained by subtracting 2 from the respective elements of the Bayer matrix 101, and the elements of the Bayer matrix 105 range from -2 to +13 inclusive. Dark areas can be expressed, and worm noise does not occur in the image.

For the second example, the image to be processed is an article of a newsletter with a big headline and illustration of plain puzzle, namely, a simple image. In Fig. 11 (a) , the pregain is set to 0.7, the random noise 201 (25%of the random noise 200) is selected, and Bayer matrix 105 is used. Some noises can be seen in flat areas. In Fig. 11 (b) , the pre gain is set to 0.6, the random noise 202 (10%of the random noise 200) is selected, and Bayer matrix 103 is used. Noises in the flat areas have been improved.

For the third example, the image to be processed is a computer graphics of natural landscape with few light and shade (not shown) , namely, a very simple image. For one case, the pre gain is set to 0.7, the random noise 201 (25%of the random noise 200) is selected, and Bayer matrix 105 is used. Some noises can be seen in flat areas. For another case, the pre gain is set to 0.6, the random noise 202 (10%of the random noise 200) is selected, and Bayer matrix 104 is used. Noises in the flat areas have been improved.

As explained above, the content adaptive control of dithering can improve quality of the flat areas on the display. If the images are shown and the flat area on the display are checked, the improvement of quality can be confirmed. In the examples of Fig. 11 (a) and Fig. 11 (b) , whole area on the display can be checked, and the improvement of quality can be confirmed.

Prior dithering techniques cause much noise such as worm noise, fixed pattern noise, etc., as described above with reference to Figs. 3 (a) and 3 (b) . In particular, Bayer matrix dithering causes fixed pattern noise having periodicity. In the embodiments of the present invention, the periodicity of the noise can be eliminated by adding random noise. However, the random noise may be noticeable in the flat areas. In the embodiment of the present invention, the amount of the random noise to be merged is adjusted based on the characteristic of the original image that can vary depending on the size of flat area. Namely, in the embodiments of the present invention, the quality of the image can be improved by balancing the added random noise with the conventional periodic noise.

Specifically, when using original images with different sizes of flat areas, the amount of noises caused by dithering differs depending on the size of flat area. When checking the processed image, the pattern caused by dithering has no periodicity, because random noise have been added.

What is disclosed above are merely exemplary embodiments of the present invention, and are certainly not intended to limit the scope of protection of the present invention. A person of ordinary skill in the art may understand that all or some of the processes that implement the foregoing embodiments and equivalent modifications made in accordance with the claims of the present invention shall fall within the scope of the present invention.

Claims

A device for dithering using a Bayer matrix, comprising:

a characteristic detection unit configured to detect a characteristic of an original image;

a control parameter unit configured to determine a control value based on the characteristic of the original image;

a first adder configured to add the control value to respective elements of the Bayer matrix to obtain modified elements of the Bayer matrix;

a second adder configured to repeatedly add a modified element of the Bayer matrix to a corresponding pixel value of the image to obtain a modified pixel value; and

a shifter configured to shift the modified pixel value to obtain an image expressed with decreased bit depth.
The device according to claim 1, wherein the characteristic of the original image is detected by using image frequency histogram.
The device according to claim 1 or 2, wherein the respective elements of the Bayer matrix are respective original elements of the Bayer matrix, or respective elements obtained by multiplying the respective original elements of the Bayer matrix by a factor that is determined based on the characteristic of the original image.
The device according to any one of claims 1 to 3, wherein the control parameter unit is further configured to determine a control gain based on the characteristic of the original image, and

the device further comprises

a first multiplier configured to multiply random noise by the control gain to obtain modified random noise; and

a third adder configured to, before repeatedly adding the modified element of the Bayer matrix to the corresponding pixel value of the image, update the respective modified elements of the Bayer matrix by adding the modified random noise to the respective modified elements of the Bayer matrix.
The device according to any one of claims 1 to 4, wherein the control parameter unit is further configured to determine a pre gain based on the characteristic of the original image, and

the device further comprises a second multiplier configured to, before repeatedly adding the modified element of the Bayer matrix to the corresponding pixel value of the image, update respective pixel values by multiplying the respective pixel values of the image by the pre gain.
The device according to any of claims 1 to 5, wherein the amount of noises, which are caused by dithering, in the image expressed with decreased bit depth differs depending on the size of flat area when using original images with different sizes of flat areas.
A device for dithering using error diffusion, comprising:

a characteristic detection unit configured to detect a characteristic of an original image;

a control parameter unit configured to determine a control value based on the characteristic of the original image;

a first adder configured to add the control value to respective coefficients for the error diffusion to obtain modified coefficients for the error diffusion;

a second adder configured to add quantization errors, which are calculated based on the modified coefficients for the error diffusion, from neighboring pixels to each pixel value of the image to obtain a modified pixel value; and

a shifter configured to shift the modified pixel value to obtain an image expressed with decreased bit depth.
The device according to claim 7, wherein the characteristic of the original image is detected by using image frequency histogram.
The device according to claim 7 or 8, wherein the respective coefficients for the error diffusion are respective original coefficients for the error diffusion, or respective coefficients obtained by multiplying the respective original coefficients for the error diffusion by a factor that is determined based on the characteristic of the original image.
The device according to any one of claims 7 to 9, wherein the control parameter unit is further configured to determine a control gain based on the characteristic of the original image, and

the device further comprises

a first multiplier configured to multiply random noise by the control gain to obtain modified random noise; and

a third adder configured to, before adding quantization errors, which are calculated based on the modified coefficients for the error diffusion, from neighboring pixels to each pixel value of the image, update the respective modified coefficients for the error diffusion by adding the modified random noise to the respective modified coefficients for the error diffusion.
The device according to any one of claims 7 to 10, wherein the control parameter unit is further configured to determine a pre gain based on the characteristic of the original image, and

the device further comprises a second multiplier configured to, before adding quantization errors, which are calculated based on the modified coefficients for the error diffusion, from neighboring pixels to each pixel value of the image, update respective pixel values by multiplying the respective pixel values of the image by the pre gain.
The device according to any of claims 7 to 11, wherein the amount of noises, which are caused by dithering, in the image expressed with decreased bit depth differs depending on the size of flat area when using original images with different sizes of flat areas.
A method for dithering using a Bayer matrix, comprising:

detecting a characteristic of an original image;

determining a control value based on the characteristic of the original image;

adding the control value to respective elements of the Bayer matrix to obtain modified elements of the Bayer matrix;

repeatedly adding a modified element of the Bayer matrix to a corresponding pixel value of the image to obtain a modified pixel value; and

shifting the modified pixel value to obtain an image expressed with decreased bit depth.
The method according to claim 13, wherein the characteristic of the original image is detected by using image frequency histogram.
The method according to claim 13 or 14, wherein the respective elements of the Bayer matrix are respective original elements of the Bayer matrix, or respective elements obtained by multiplying the respective original elements of the Bayer matrix by a factor that is determined based on the characteristic of the original image.
The method according to any one of claims 13 to 15, further comprising:

determining a control gain based on the characteristic of the original image;

multiplying random noise by the control gain to obtain modified random noise; and

before repeatedly adding the modified element of the Bayer matrix to the corresponding pixel value of the image, updating the respective modified elements of the Bayer matrix by adding the modified random noise to the respective modified elements of Bayer matrix.
The method according to any one of claims 13 to 16, further comprising:

determining a pre gain based on the characteristic of the original image; and

before repeatedly adding the modified element of the Bayer matrix to the corresponding pixel value of the image, updating respective pixel values by multiplying the respective pixel values of the image by the pre gain.
The device according to any of claims 13 to 17, wherein the amount of noises, which are caused by dithering, in the image expressed with decreased bit depth differs depending on the size of flat area when using original images with different sizes of flat areas.
A method for dithering using error diffusion, comprising:

detecting a characteristic of an original image;

determining a control value based on the characteristic of the original image;

adding the control value to respective coefficients for the error diffusion to obtain modified coefficients for the error diffusion;

adding quantization errors, which are calculated based on the modified coefficients for the error diffusion, from neighboring pixels to each pixel value of the image to obtain a modified pixel value; and

shifting the modified pixel value to obtain an image expressed with decreased bit depth.
The method according to claim 19, wherein the characteristic of the original image is detected by using image frequency histogram.
The method according to claim 19 or 20, wherein the respective coefficients for the error diffusion are respective original coefficients for the error diffusion, or respective coefficients obtained by multiplying the respective original coefficients for the error diffusion by a factor that is determined based on the characteristic of the original image.
The method according to any one of claims 19 to 21, further comprising:

determining a control gain based on the characteristic of the original image;

multiplying random noise by the control gain to obtain modified random noise; and

before adding quantization errors, which are calculated based on the modified coefficients for the error diffusion, from neighboring pixels to each pixel value of the image, updating the respective modified coefficients for the error diffusion by adding the modified random noise to the respective modified coefficients for the error diffusion.
The method according to any one of claims 19 to 22, further comprising:

determining a pre gain based on the characteristic of the original image; and

before adding quantization errors, which are calculated based on the modified coefficients for the error diffusion, from neighboring pixels to each pixel value of the image, updating respective pixel values by multiplying the respective pixel values of the image by the pre gain.
The device according to any of claims 19 to 23, wherein the amount of noises, which are caused by dithering, in the image expressed with decreased bit depth differs depending on the size of flat area when using original images with different sizes of flat areas.
A device for dithering using a Bayer matrix, comprising a processor and a memory, wherein the processor is coupled to the memory, the memory is configured to store a computer program, and the processor is configured to execute the computer program in the memory, to perform the method according to any one of claims 13 to 18.
A device for dithering using a Bayer matrix, comprising a processor and a memory, wherein the processor is coupled to the memory, the memory is configured to store a computer program, and the processor is configured to execute the computer program in the memory, to perform the method according to any one of claims 19 to 24.