KR20070022596A - Method and device for processing video data to be displayed on a display device - Google Patents

Abstract

The present invention relates to a method for processing video data to be displayed on a display device having a plurality of light emitting elements corresponding to pixels of a video picture, wherein the error diffusion step refines the gray scale description of the video picture. Applied to at least a portion of the video data, wherein the error spreading step comprises truncating a value of the corresponding video data and truncating error for each current pixel of the portion of the video picture. Diffusing to adjacent pixels of. According to the present invention, noise is inserted into the error before and / or after spreading to the adjacent cell. Because of this principle, no static pattern should be seen while improving the overall image quality.

Description

METHOD AND DEVICE FOR PROCESSING VIDEO DATA TO BE DISPLAYED ON A DISPLAY DEVICE}

1 illustrates the basic steps of a prior art error diffusion process.

2 is a block diagram of an error diffusion circuit of the prior art;

3 is a block diagram of a first embodiment of an error diffusion circuit according to the present invention;

4 is a block diagram of a second embodiment of an error diffusion circuit according to the present invention;

5 is a block diagram of a chain random noise insertion circuit of the error spreading circuit of FIG. 3 or FIG.

6 is a block diagram of a bit noise adder of the chain random noise insertion circuit of FIG.

7 is a flow chart describing the operation of the bit noise adder of FIG.

8 is a block diagram of a noise insertion circuit that may be used in the error spreading circuit of FIG. 3 or 4 instead of the chain random noise insertion circuit of FIG.

<Explanation of symbols for the main parts of the drawings>

10: summing circuit 20: cutting circuit

30: diffusion calculation circuit 40: chain random noise insertion circuit

40 ': noise insertion circuit 41: random generator

42: subtraction circuit 43: minimum value transfer circuit

44: multiplication circuit 45: multiplication circuit

46: addition circuit 60: first AND gate

61: second AND gate 62: NOT gate

63: first OR gate 64: second OR gate

65: third AND gate BNA [i]: noise adder

The present invention relates to a method for processing video data to be displayed on a display device having a plurality of light emitting elements corresponding to pixels of a video picture, wherein the error diffusion step refines the gray scale depiction of the video picture. applied to at least a portion of the video data for refinement, the error spreading step truncating the truncation error and truncating the value of the corresponding video data for each current pixel of the portion of the video picture. Diffusing to at least one adjacent pixel. The invention also relates to a device implementing the method.

Plasma display panels, called PDPs, utilize a matrix array of discharge cells, which can have only two states of "on" or "off". Also unlike CRT or LCT screens where the gray level is represented by analog control of light emission, the PDP controls the gray level by modulating the number of light pulses (hold pulses) per frame. This temporal modulation will be integrated by the eye in a period corresponding to the eye time response. Since the video amplitude is depicted by the number of light pulses occurring at a given frequency, while at a given frequency, a higher amplitude means more light pulses, and therefore longer time when the cell is "on". For this reason, this kind of modulation is also known as Pulse Width Modulation (PWM).

This PWM is responsible for one of the PDP image quality issues, i.e. poor gray scale portrayal quality, especially in dark areas of the image. This is due to the fact that the displayed luminescence is linear with the number of pulses, but the eye's response and sensitivity to noise are not linear. In darker areas, the eye is more sensitive in brighter areas. This means that although modern PDPs can display 255 discrete video levels, quantization errors are very noticeable in the dark areas of the picture.

To achieve a better gray scale depiction, the dither signal is typically added to the video signal before truncation to the final video gray scale amplitude resolution. Dithering is a well known technique used to reduce the effects of quantization noise due to the reduced number of displayed resolution bits. Dithering improves gray scale depiction by adding artificial levels in between, but also adds high frequency low amplitude dithering noise that can be perceived by human observers at close viewing distances.

Currently, there are mainly two types of dithering used for PDPs:

Cell-based dithering disclosed in European patent application EP 1 269 457: This improves gray scale description but adds high frequency low amplitude dithering noise, as mentioned previously. When no movement is displayed, dithering is fully integrated by the eye, and the dithering pattern is not noticeable; In this case, the dithering pattern is hidden by spatial and temporal eye integration; However, while displaying motion in the picture, the dithering pattern can become more pronounced.

Error-diffusion: This is based on the distribution of the gray level fragment of a given cell to adjacent cells, improves gray scale depiction, and produces no dithering pattern. This mainly adds noise in the dark areas, but this noise looks more natural than the pattern of cell-based dithering. However, for a static picture without temporal noise, this creates a very visible static pattern.

It is an object of the present invention to disclose a method and apparatus for removing the static pattern of error diffusion. This is accomplished by inserting noise for truncation errors to be spread in adjacent cells, the amplitude of which depends on the value of the truncation error. Because of this principle, no static pattern should be visible while improving the overall picture quality.

According to the invention, this object is solved by a method for processing video data to be displayed on a display device having a plurality of light emitting elements corresponding to pixels of a video picture, wherein the error diffusion step is said part of the video picture. Is applied to at least a portion of the video data to refine the gray scale depiction of the. The error spreading step includes truncating a value of corresponding video data for each current pixel of the portion of the video picture and spreading the truncation error to at least one adjacent pixel. According to the invention, the method comprises inserting noise into the truncation error, the amplitude of the noise being inserted depends on the value of the truncation error to be spread. Since the inserted noise should not change the average value of its output, the noise has an average value of zero. So if the amplitude of the noise is A, then the noise is a randomly chosen value between -A and A. The inserted noise is, for example, white noise.

If a truncation error is spread over a plurality of adjacent pixels, spreading the truncation error includes, for example, extending a spread value from the truncation error for each of the adjacent pixels, The sum of is approximately equal to the truncation error. In the first embodiment, noise is added to the truncation error before the calculating step. In the second embodiment, the insertion of noise into truncation error is performed after the calculating step and inserts noise into each spreading value.

The invention also relates to a device for processing video data to be displayed on a display device having a plurality of light emitting elements corresponding to pixels of a video picture, which device refines the gray scale depiction of this portion of the video picture. Error spreading circuitry for applying error spreading to at least a portion of the video data. An error diffusion circuit includes a truncation circuit for truncating the value of the corresponding video data for each current pixel of the portion of the video picture, and for each of the adjacent pixels, a diffusion for calculating a spread value from the truncation error. As a calculation circuit, a sum of diffusion values for the adjacent pixels includes a diffusion calculation circuit, which is approximately equal to the truncation error, and a summation circuit for adding the corresponding diffusion values to each adjacent pixel. According to the invention, the device further comprises a noise insertion circuit for inserting white noise into the truncation error before and / or after this error is spread to adjacent pixels.

In the first embodiment, the noise insertion circuit is located between the cutting circuit and the diffusion calculating circuit. In the second embodiment, the noise insertion circuit is located between the spreading calculation circuit and the summation circuit to insert noise into the spread value.

By way of example, the noise insertion circuit includes a random generator for generating random data and a plurality of series-connected bit noise adders for inserting noise into the truncation error, each bit noise adder being associated with the random data. Used to generate at least one bit of the noise truncation error from the truncation error. Preferably, a bypass signal is provided independently for each bit noise adder to bypass the bit noise adder to adapt the noise level.

In another example, the noise insertion circuitry comprises a random generator for generating random numbers between -1 and +1, multiplying a constant random number by a function of the truncation error, and adding this multiplied value to the truncation error. It includes a circuit.

Exemplary embodiments of the invention are illustrated in the drawings and described in more detail in the following description.

As mentioned previously, the PDP utilizes pulse width modulation (PWM) to generate grays of different shades. In contrast to CRT screens, which are quadratic approximations to the cathode voltage to which luminescence is applied, luminescence is linear with the number of discharge pulses. Therefore, approximately a digital second order degamma function should be applied to the video signal prior to pulse width modulation.

Because of this degamma function, for smaller video levels, many input levels are mapped to the same output level. In other words, in the darker regions, the output number of quantization bits is smaller for input numbers, especially for values less than 16 that all map to 0 (when operating with 8 bitwords for the input video signal and gray level). . This corresponds to a loss of 4-bit resolution that is practically unacceptable for video applications.

Dithering is a known technique for avoiding the loss of amplitude resolution bits by truncation of the output level after the degamma function. This only works if the required resolution is available before truncation, that is, more bits are used for the degamma function. Dithering can in principle recover as many bits as bits lost by truncation.

1-bit dithering corresponds to multiplying the number of available output levels by 2,

2-bit dithering corresponds to multiplying the number of available output levels by 4,

3-bit dithering corresponds to multiplying the number of available output levels by 8, and in other cases this ratio also follows.

Error diffusion is a neighboring operation that quantizes the input value of the current pixel (maintains an integer part of the input value in the present case), and sends truncation errors (fractional parts) to future pixels. Officially, Floyd and Steinberg {"An adaptive algorithm for spatial greyscale" in Proc. Soc Information Display, 1976, vol 17, no. 2, pp. 75-78}

이 되도록 입력 픽셀 x[n]을 조정하고 반올림하는 것에 의해 출력 픽셀 y[n]을 한정하는데, 여기서

은 이전 반복 동안

로서 축적된 확산 에러(현재 경우에서는 소수 부분)이고, 여기서,

은

인 다양한 소수 부분을 나타낸다.

To define the output pixel y [n] by adjusting and rounding the input pixel x [n] to

During the previous iteration

Is a spreading error (in the present case, a fractional part), where

silver

Represents various fractional parts.

The error diffusion process itself consists of three steps. Each input value is the sum of the original input value and the spread past error computed for the adjacent pixel, which is, for example, a pixel located above and below the current pixel. In a first step, this input value is rounded to produce an output, and truncation error (fractional part) is defined as the difference between the rounded input value and the initial input value. In the next step, this truncation error spreads to adjacent pixels that follow a coefficient that can be selected in various ways. Then, the next pixel (e.g., the pixel located to the right of the current pixel) is processed in the same way.

1 illustrates this process. At the beginning of the first step, the value of the current pixel is 4.5. This value is then rounded to 4 creating a truncation error of 0.5. This truncation error is spread to three adjacent pixels using three different diffusion coefficients (0.5 for the right pixel, 0.3 for the bottom right pixel, 0.2 for the bottom pixel). The coefficients are selected to maintain an energy constant (sum of the coefficients is 1). This is essential to maintain good stability in the image. After this step, the three pixel values are:

오른 쪽 픽셀에 대해 4.85 + 0.5 ×0.5 = 5.1;

4.85 + 0.5 x 0.5 = 5.1 for the right pixel;

하단 오른 쪽 픽셀에 대해 4.55 + 0.5 ×0.3 = 4.7;

4.55 + 0.5 × 0.3 = 4.7 for the lower right pixel;

하단 픽셀에 대해 5.10 + 0.5 ×0.2 = 5.2로 변경된다.

It is changed to 5.10 + 0.5 x 0.2 = 5.2 for the bottom pixel.

This process then applies to the right pixel with a value of 5.1, which becomes the current pixel.

The basic error diffusion circuit is shown in FIG. This circuit is provided for each color element (red, green and blue). Summing circuit 10 implements the sum of the original input value coming from input video signal Y_IN [14: 0] and the spread past error coming from spreading calculation circuit 30. The truncation circuit 20 rounds the input value (integer portion of the input value) coming from the circuit 10 and generates an output value Y_OUT [7: 0] and a truncation error (fractional portion of the input value). The diffusion calculation circuit 30 truncates the diffusion value for each adjacent pixel (one pixel for the right pixel, one pixel for the bottom right pixel, one for the bottom pixel in the current example), as described above. Calculate from Since the input value is input to summing circuit 10, this spreading value is added to the original input value of this adjacent pixel.

In this example, the gray level after the degamma function (Y_IN [14: 0]) is coded 15 bits, i.e. 8 bits for the integer part and 7 bits for the fractional part, and the gray level after dithering is 8 bits (integer). Part).

Although the error-diffusing image is good to see (the introduced noise is similar to natural video noise), this algorithm produces some unwanted textures for the static image.

One idea to attenuate the static pattern of error diffusion is to randomly change the diffusion coefficient. Unfortunately, since this means changing the internal structure of the error spreading network 30, this is only possible for some basic implementation of error spreading.

According to the invention, it is proposed to add noise to the truncation error, the amplitude of which is dependent on the value of the truncation error. This may be implemented by a specific circuit called a chain random noise insertion circuit, which may be inserted between circuit 20 and circuit 30 or between circuit 30 and circuit 10. Since this only changes the error value to be spread, it does not require modification of other circuitry and can probably be used with any error spreading algorithm. The inserted noise should not change the average value of the output, resulting in noise having an average value of zero. So if the amplitude of the noise is A, this noise is a randomly chosen value between -A and A. This noise must also be dependent on the value of the truncation error, so that its amplitude is small when the truncation error is small (eg, when the truncation error is zero, no noise should be added). This noise may be white noise, for example.

3 shows a first embodiment of an error diffusion circuit according to the present invention. A chain random noise insertion circuit 40 is inserted between the truncation circuit 20 and the diffusion calculation circuit 30. This adds white noise to the error provided to the spreading calculation circuit 30.

4 illustrates a second embodiment of the error diffusion circuit according to the present invention. A chain random noise insertion circuit 40 is inserted between the spreading calculation circuit 30 and the summing circuit 10. This circuit adds white noise to the spread value provided to the summation circuit 10.

In another possible embodiment, noise may be inserted before and after truncation error spreading. Two concatenated random noise insertion circuits will be used, one before and one after the spreading calculation circuit 30.

One example of a chain random noise insertion circuit 40 that receives an input signal in [6: 0] and generates an output signal out [6: 0] is illustrated in FIG. 5. This circuit recursively introduces noise from the most significant bit (in [6]) of the input value (in [6: 0]) to the least significant bit (in [0]) of the input value. It works by adding Since this circuit consists of a six-bit noise adder BNA [i] operating in series, i.e. one output S [i] of each bit noise adder BNA [i] is a bit noise adder BNA. This circuit is called a chain random noise insertion circuit. It also includes a random generator 50 that outputs six independent random numbers r [0] to r [5].

Each bit noise adder BNA [i] may be independently deactivated by a bypass signal bypass [i]. This may be useful to minimize noise levels. For example, if it is desirable to add noise only on the five least significant bits (instead of seven), bypass [5] and bypass [4] should be set to 1 and bypass [3] to bypass [0]. Must be set to zero. If bypass [i] = 0 to have bypass [j] = 1 for j <i, this does not mean anything. Indeed, it is possible to use this bypass to reduce the value of noise added by the present invention. As indicated previously, the amplitude of this noise is approximately equal to the error value with a maximum of one half. This maximum value can be reduced to 1/4 by having bypass [5: 0] = 100000, and can be reduced to 1/8 by having bypass [5: 0] = 110000.

More specifically, the signal processed by the bit noise adder is as follows:

The bit noise adder BNA [5] receives the signals in [6], in [5], bypass [5] and r [5] and carries the signals out [6] and s [5]. do;

The bit noise adder BNA [4] receives the signals in [4], s [5], bypass [4] and r [4] and carries the signals out [5] and s [4]. do;

The bit noise adder BNA [3] receives the signals in [3], s [4], bypass [3] and r [3] and carries the signals out [4] and s [3]. do;

The bit noise adder BNA [2] receives the signals in [2], s [3], bypass [2] and r [2] and carries the signals out [3] and s [2]. do;

The bit noise adder BNA [1] receives the signals in [1], s [2], bypass [1] and r [1] and carries the signals out [2] and s [1]. do;

The bit noise adder BNA [0] receives the signals {in [0] (hereinafter referred to as s [0]), s [1], bypass [0] and r [0]) and outputs the signal out [ 0] and out [1]).

Each bit noise adder BNA [i] has the same structure. One example of the structure of the bit noise adder BNA [0] is given in FIG. 6. The bit noise adder is:

A first AND gate 60 which receives signals {in [0] (= s [0]) and r [0]} as input,

NOT gate 62 for negating the signal bypass [0],

A second AND gate 61 which receives as its input the output signal of the AND gate 60 and the negated bypass [0],

A first OR gate 63 which receives the output signal of the AND gate 61 and the signal s [1] as input and carries the signal out [1],

A second OR gate 64 which receives signals s [1] and bypass [0] as input;

A third AND gate 65 which receives as input the output signal of the OR gate 64 and the signal in [0] and carries the signal out [0].

The function of this structure can be summarized by FIG. This block BNA [0] delivers an output signal different from the input signal only when s [1] = 0 and s [0] = 1. In this case, the output signal out [1: 0] is set to 00 or 10, depending on the value of the random number r [0]. When r [0] = 1, out [1: 0] is set to 10, and when r [0] = 0, out [1: 0] is set to 00. Thus, if s [1: 0] is equal to 01, the output signal out [1: 0] is also equal to 01 on average. This means that the output signal out [1: 0] is on average equal to the value of the signal s [1: 0]. The energy of the error is thus kept constant throughout the picture. The only values of truncation error (in [6: 0]) where no noise is added are 0000000, 1000000, 1100000, 1110000, 1111000, 1111100, 1111110 and 1111111. When this truncation error is equal to zero, it is important not to add any noise to the truncation error (this means within one region where no dithering is required). For other values where no noise is added (1000000, 1100000, 1110000, 1111000, 1111100, 1111110 and 1111111), the adjacent area will in fact have different values (because of the propagation of the error) and all areas where the truncation error is not zero. Some noise is also added globally in.

It is also important to note that when the signal s [1: 0] is equal to 00, the output signal out [1: 0] is also equal to 00; This means that this block adds no noise in the black areas of the picture.

Below, the results of the random noise insertion circuit for three examples can be found.

First example : in [6: 0] = 0011010 and r [5: 0] = 101001

in [6] = 0 and in [5] = 0 and r [5] = 1, so out [6] = 0 and s [5] = 0

in [4] = 1 and s [5] = 0 and r [4] = 0, so out [5] = 0 and s [4] = 0

in [3] = 1 and s [4] = 0 and r [3] = 1, so out [4] = 1 and s [3] = 0

in [2] = 0 and s [3] = 0 and r [2] = 0, so out [3] = 0 and s [2] = 0

in [1] = 1 and s [2] = 0 and r [1] = 0, so out [2] = 0 and s [1] = 0

s [0] = 0 and s [1] = 0 and r [0] = 1, so out [1] = 0 and out [0] = 0. So finally out [6: 0] = 0010000.

Second example : in [6: 0] = 1010001 and r [5: 0] = 101001

in [6] = 1 and in [5] = 0 and r [5] = 1, so out [6] = 1 and s [5] = 0

in [4] = 1 and s [5] = 0 and r [4] = 0, so out [5] = 0 and s [4] = 0

in [3] = 0 and s [4] = 0 and r [3] = 1, so out [4] = 0 and s [3] = 0

in [2] = 0 and s [3] = 0 and r [2] = 0, so out [3] = 0 and s [2] = 0

in [1] = 1 and s [2] = 0 and r [1] = 0, so out [2] = 0 and s [1] = 0

s [0] = in [0] = 1 and s [1] = 0 and r [0] = 1, so out [1] = 1 and out [0] = 0. So finally out [6: 0] = 1000010.

Third example : in [6: 0] = 1110100 and r [5: 0] = 101001

in [6] = 1 and in [5] = 1 and r [5] = 1, so out [6] = 1 and s [5] = 1

in [4] = 1 and s [5] = 1 and r [4] = 0, so out [5] = 1 and s [4] = 1

in [3] = 0 and s [4] = 1 and r [3] = 1, so out [4] = 1 and s [3] = 0

in [2] = 1 and s [3] = 0 and r [2] = 0, so out [3] = 0 and s [2] = 0

in [1] = 0 and s [2] = 0 and r [1] = 0, so out [2] = 0 and s [1] = 0

s [0] = in [0] = 0 and s [1] = 0 and r [0] = 1, so out [1] = 0 and out [0] = 0. So finally out [6: 0] = 1110000.

Preferably, in the final circuit implementation, different random outputs r [5: 0] will be used for different color elements R, G and B. If the same random output (r [5: 0]) is used for different color elements, the error and therefore the dither structure is the same for these three elements, so dithering can be more visible (white pixels are separated Is more visible than red, green and blue), which is particularly beneficial in the case of monochrome signals.

If not, the random generator 41 should provide a new output signal for each frame, or preferably for each pixel.

Instead of using the chain random noise insertion circuit 40 described in FIG. 5, it is possible to use the noise insertion circuit 40 ′ described in FIG. 8: This noise insertion circuit 40 ′ is:

A random generator 41 which carries a random number r [6: 0] and a random number code bit sr [0] corresponding to the sign of this random number;

A subtraction circuit 42 for subtracting the truncation error in [6: 0] from this value 1;

A circuit for conveying the minimum value between the truncation error (in [6: 0]) and the output signal of the subtraction circuit 42 (as a result of the subsequent addition, the output signal is still contained between 0 and 1). (43);

A multiplication circuit 44 for multiplying the output signal of the circuit 43 by the random number r [6: 0];

A multiplication circuit 45 for multiplying the output signal of the circuit 44 by the random number bit sr [0], and

An addition circuit 46 for adding together the truncation error in [6: 0] and the output signal of the circuit 45.

This noise insertion circuit has similar characteristics to chain random noise insertion:

The added insert has an average value equal to zero.

This noise is also dependent on the truncation error value, so that when the truncation error is small, its amplitude becomes small (when the truncation error is equal to 0, this circuit adds no noise to the truncation error).

The invention likewise applies to any type of noise insertion circuit with the above mentioned characteristics.

The invention is applicable to a method for processing video data to be displayed on a display device having a plurality of light emitting elements corresponding to pixels of a video picture.

Claims (10)

A method for processing video data to be displayed on a display device having a plurality of light emitting elements corresponding to pixels of a video picture, the method comprising: 18. A method for processing video data wherein an error diffusion step is applied to the at least a portion of the video data to refine a gray scale portrayal of at least a portion of a video picture, The error spreading step includes truncating the value of the corresponding video data for each current pixel of the portion of the video picture and diffusing the truncation error resulting from the truncation step into at least one adjacent pixel. In the method comprising: The method further comprises inserting noise into the truncation error, wherein the amplitude of the noise to be inserted is dependent upon the value of the truncation error to be spread. The method of claim 1, wherein the inserted noise is white noise. 3. The method of claim 1 or 2, wherein if the truncation error is spread across a plurality of adjacent pixels, spreading the truncation error includes extending a spread value from the truncation error for each of the adjacent pixels. And wherein the sum of the spread values for the adjacent pixels is substantially equal to the truncation error, and the noise is inserted into the truncation error prior to the calculating step. 3. The method of claim 1, wherein if the truncation error is spread over a plurality of adjacent pixels, spreading the truncation error includes calculating a spread value from the truncation error for each of the adjacent pixels; And the sum of the spread values for the adjacent pixels is substantially equal to the truncation error, and the insertion of the noise into the truncation error inserts noise into each spread value. A device for processing video data to be displayed on a display device having a plurality of light emitting elements corresponding to pixels of a video picture, the device comprising: An error diffusion circuit for applying error diffusion to at least a portion of the video data to refine the gray scale representation of at least a portion of the video picture, The error diffusion circuit Cutting circuit 20 for truncating the value of the corresponding video data, for each current pixel of the portion of the video picture, For each of at least one adjacent pixel of the current pixel, a diffusion calculation circuit 30 for calculating a diffusion value from the truncation error, wherein the sum of the diffusion values for the adjacent pixels is approximately equal to the truncation error. Diffusion calculation circuit 30, A summation circuit (10) for adding the corresponding spreading value to each adjacent pixel, the device for processing video data, comprising: And a noise insertion circuit (40) for inserting noise into the truncation error, wherein the amplitude of the inserted noise is dependent on the value of the truncation error to be spread. 6. Device according to claim 5, characterized in that the noise insertion circuit (40) inserts white noise into the truncation error before or after spreading. Device according to claim 5 or 6, characterized in that the noise insertion circuit (40) is located between the truncation circuit (20) and the diffusion calculation circuit (30). 7. The noise inserting circuit (40) according to claim 5 or 6, characterized in that the noise inserting circuit (40) is located between the spread calculating circuit (30) and the summing circuit (10) to insert noise into the spread value. Device for processing video data. The noise inserting circuit (40) of claim 5, wherein the noise inserting circuit (40) comprises a random generator (50) for generating random data and a plurality of series-connected bit noises for inserting noise into the truncation error. An adder (BNA [i]), each bit noise adder being used to generate at least one bit of said noise truncation error from said random data and said truncation error. device. 10. The video of claim 9, wherein a bypass signal is provided independently to each bit noise adder (BNA [i]) to bypass the bit noise adder to adapt the noise level. Device for processing data.

Images