JP3305837B2 - Anti-aliasing method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to anti-aliasing in generating images in computer graphics (CG).

[0002]

2. Description of the Related Art Sampling of a space (sam) such as ray tracing (ray tracing, line-of-sight search).
When an image is generated by pling (sample extraction), the image may be jagged. This jaggedness is called aliasing (edge distortion). This is roughly divided into the following two phenomena (a) and (b). (a) A phenomenon in which the edge (edge) becomes step-like without being smooth. This is called jaggy. One of the edge portions is a boundary portion of the object. For example, the boundary between the object and another object, the object and the background, and the same object, such as the side surface and the upper surface, becomes a problem. The other is a case where the object has a pattern. In this case, the pattern itself is the boundary, and the edge of the pattern becomes a problem even if it is not the edge of the object. In the case of a classic checker-like mapping, the jaggedness is noticeable. The same applies to the case of reflection, transmission, refraction, and shadow. (b) A phenomenon in which a small object or part of a large object disappears. This is an example of a special edge, but whether it is a still image or an animation, the frames can make the object visible or invisible depending on the frame.

[0003] In order to avoid such jaggedness, an operation called anti-aliasing is performed. The most well-known method is that the edge is a portion where the color changes drastically, and the jagged portion there is conspicuous, so increase the sampling of that portion (this is called oversampling (oversampling)). And there are ways to avoid jaggedness. The edge recognition is performed when the color change value exceeds (or exceeds) a given fixed threshold value (threshold). However, anti-aliasing is very expensive and requires a lot of computer resources, especially
Consumes CPU resources. It is not uncommon for the anti-aliasing to be several times to ten and several times more. Therefore, we need to improve the recognition of edges and improve efficiency. If the edge recognition is the worst, it is the same as generating an image with several times the target number of pixels and simply reducing it to the target number of pixels.

As an example of the technique, there is a method of recognizing whether or not there is a difference between sampled objects, instead of recognizing an edge based on a change in color. However, this method does not perform anti-aliasing when it is mapped to an object. This is because even if they are mapped in this method, oversampling is not performed as long as they are the same object.

[0005]

The recognition of an edge is ultimately to be seen by the human eye as a jagged edge, and the object does not matter. Given the characteristics of the human eye, it should be possible to recognize edges more accurately and increase efficiency. The present invention pays attention to this point and aims to provide an efficient anti-aliasing method.

Here, terms and related matters will be explained.

Generally, CG handles three primary colors of red (Red), green (Green), and blue (Blue), and each range of the three primary colors is 0 or more.
Generally, it is expressed by a real value of 1 or less. On the other hand, when finally writing to a frame buffer or a file, it is within the scope of the processing system. For example, R
In the process of managing GB by 8 bits, it becomes an integer value of 0 or more and 255 or less. In reality, RGB is 8 bits each, 24 bits in total, that is, 24 bits
There is a processing system of bit / dot or more, and the most versatility is based on the real value of 0-1. Therefore, in this specification, the range of the three primary colors is represented by a real value of 0 or more and 1 or less. In addition, 0-1
The value below a certain value is meaningless even if it is said to be a real value of. If the target processing system is 24bit / dot, the limit of the step is 1/255
$ 0.0039.

When a color is represented by an RGB display system rgb (red, green, blue)
And For example, rgb (1,0,0) for red, rgb for green
(0,1,0), blue for rgb (0,0,1), white for rgb (1,1,1)
It is. rgb (0.18,0.18,0.18) indicates gray with a reflectance of 18%.

The "difference between A and B" indicates an absolute value. That is, the difference between A and B = | A-B | or the difference between A and B = | B-A |.

[0010]

Claims 1 and 2 are inventions of a method for determining a threshold value. The targeted anti-aliasing method compares each of the red, green, and blue values at the two sample points separately, and any one of the differences exceeds each of the preset thresholds (or thresholds). In this case, over-sampling is performed, which is the most well-known anti-aliasing method. In this case, if the range of the three primary colors is 0 or more and 1 or less, the range of the threshold value is also 0 or more and 1 or less. Anti-aliasing is performed more carefully as the threshold value is smaller.

There are the following meaning differences between "when the threshold is exceeded" and "when the threshold is exceeded". If the threshold is exceeded, no oversampling is performed if the threshold is 1, and if the threshold is 0, oversampling is performed unless the difference in color change is 0. When the threshold value is negative, oversampling is performed even when the difference in color change is 0. In the case of "above the threshold", if the threshold is 1, oversampling is performed only when the difference in color change is 1, and if the threshold is 0, oversampling is performed even if the difference in color change is 0. Sampling is performed. If the threshold exceeds 1, no oversampling is performed.

The superiority of both "when the threshold value is exceeded" and "when the threshold value is exceeded" is not an object of the present invention, but the effect is clearer when "the threshold value is exceeded". The reason is that a negative threshold value may be specified in the case of a special condition that oversampling is performed even when the difference in color change is 0, that is, when oversampling is unnecessary. The threshold 1 does not cause any oversampling, and the threshold 0 requires that oversampling be performed whenever necessary.

The two inventions relate to a method of determining a threshold value. The invention of claim 1 sets a green threshold value as a reference, and the red and blue threshold values are based on the green threshold value. The threshold value for red is larger than the threshold value for green, and the threshold value for blue is larger than the threshold value for red. Claim 2
The invention of (1) is a specific example of the invention of claim 1, characterized in that a red threshold is a green threshold 0.587 / 0.299, and a blue threshold is a green threshold 0.587 / 0.114.

According to a third aspect of the present invention, in the anti-aliasing method based on oversampling, the color of two sample points is represented by a luminance component indicating brightness, a chromaticity component having high human eye discrimination ability, and a human eye. Is divided into three chromaticity components having low discrimination ability, and each of the three components is compared separately, and any one of the respective differences exceeds (or is equal to or greater than or equal to) the predetermined threshold value. An anti-aliasing method is characterized in that oversampling is performed in such a case.

However, the three preset threshold values are increased in the order of a luminance component indicating brightness, a chromaticity component having high human eye discrimination ability, and a chromaticity component having low human eye discrimination ability.

The invention according to claim 4 is a specific example of the invention according to claim 3. One threshold is set, and this is set as the threshold of the Y component. The threshold of the I component is the threshold of the Y component, the bandwidth of the Y signal / the bandwidth of the I signal, and the threshold of the Q component is the threshold of the Y component. Bandwidth /
Three thresholds are determined as the bandwidth of the Q signal, the colors of the two sample points are converted from the RGB display system to the YIQ display system, and each of the three components of the YIQ is compared separately. One is an anti-aliasing method in which oversampling is performed when each threshold value is exceeded (or is equal to or greater than the threshold value).

According to the fifth and sixth aspects of the present invention, the chromaticity component having a low human eye discrimination ability is ignored from the colors of the two sample points, and the luminance component indicating the brightness and the chromaticity having the high human eye discrimination ability are ignored. Taking two of the components, comparing each of the two components separately, and oversampling if any one of the differences exceeds (or exceeds) a preset threshold. The anti-aliasing method is characterized in that:

However, the invention of claim 6 is characterized in that the threshold value of the luminance component indicating the brightness is made smaller than the threshold value of the chromaticity component having high human eye discrimination ability.

The invention of claim 7 is a specific example of the invention of claim 6. One threshold is set, this is set as the threshold of the Y component, the threshold of the I component is set as the threshold of the Y component, 4.2 / 1.5, and two thresholds are determined. From the colors of the two sample points, Y and I of the YIQ display system are determined.
And comparing each of the Y and I components separately, and oversampling when any one of the differences between Y and I exceeds (or exceeds) the respective threshold. This is an anti-aliasing method.

Incidentally, the invention of the anti-aliasing method includes the following H1 and H2. [H1] In the anti-aliasing method by oversampling, the colors of two sample points can be converted from RGB display system to ULC
An anti-aliasing method comprising converting to an S display system and performing over-sampling when the color difference exceeds (or exceeds) a preset threshold.

[H2] In the anti-aliasing method based on oversampling, the color difference between two sample points is calculated by the color difference formula in NBS unit, the color difference formula by U * V * W * display system, and the L * a * b * display system. Color difference formula, L * u * v * display system color difference formula, Adams-Kickerson color difference formula, Hunter color difference formula, and the color difference exceeds a preset threshold value (or threshold value). An anti-aliasing method characterized in that oversampling is performed in the above case).

[0022]

The function of the invention as defined in claims 1 and 2. There are three threshold values for the three primary colors, and the prior art has no way to handle it. FIG. 2 shows a conventional example thereof (← in the equation in the figure indicates substitution). At present, the user has no choice but to give the same threshold value to all three. Some processing systems perform this on the system side. Therefore, the present invention focused on the color difference discrimination ability of the human eye. Human eyes differ in the ability to discriminate between the three primary colors, with green having the highest discrimination ability, followed by red and blue.

According to the first aspect of the present invention, it is only necessary to set a threshold based on green having a high color difference discrimination ability, and the red and blue thresholds are weighted based on the green threshold in accordance with the color difference discrimination ability of the human eye. Is determined.

CG is generally viewed on a television receiver. NTSC (National Television System Com
mittee) color broadcasts use a luminance signal (Y signal). The black and white TV uses this. Therefore, it is a practical method to determine the threshold value based on this ratio. This is the second aspect of the present invention. Green threshold This is because green has the highest discrimination ability. If you refer to blue Even if the blue threshold is 1, the green threshold is 0.194, which is a bad result. Input threshold range is 0 or more
Although it is one method to set the threshold to 5.15 or less, the purpose is to simplify determination of the threshold, which is not preferable.

On the other hand, in a high-definition television (HDTV), a luminance signal (Ey signal) is And When a color image is converted into a black-and-white image according to (Equation 4), an impression that the skin color is slightly darker is obtained as compared with the image converted according to (Equation 1). (Equation 2) is Becomes R: G: B = 3: 6: 1 for NTSC, and R: G: B = 2: 7: 1 for HDTV. (Equation 2) and
(Equation 5) has a considerable difference in the numerical values, but the basic effect remains unchanged.

The operation of the invention according to claims 3 to 7. These inventions (especially claims 4 and 7) are based on the NTI system YIQ
A more aggressive use of transformations. The human eye tends to look less saturated when the color area is smaller.
As the area of the color is reduced, the color on the chromaticity diagram becomes a color along the axis connecting orange and cyan. Green and blue cannot be distinguished at a viewing angle of 20 'and yellow and blue at a viewing angle of 2'.
It is said that when it is 1 ', it becomes achromatic (1949 Middleton (WEKMiddletone) and Holmes (MCHolmes)
Experiment). In other words, very fine things are brightness (brightness)
Just distinguish them.

Based on these eye characteristics, the NTSC system sets the luminance Y to a wide bandwidth, in phase with the orange-cyan axis (Inp).
The efficiency is increased by transmitting the I-axis in hase) with a rather narrow bandwidth and the quadrature phase (Q-axis (yellow-magenta axis)) with a very narrow bandwidth. It is not necessary to give the same amount of information to each of the three primary colors and RGB, and this was done in order to use limited resources such as radio waves effectively. On the other hand, the present invention seeks to utilize the characteristics of the eyes for effective use of computer resources.

Although the RGB display system can be converted to the YIQ display system by the chromaticity equation of (Equation 3), there is no meaning in itself, and it is important that the bandwidth, that is, the information amount is good if Y>I> Q. It is. The threshold will be weighted with this bandwidth. Y is a luminance signal component, I is a broadband chromaticity signal component, and Q is a narrowband chromaticity signal component. In the NTSC system, the bandwidth is Y signal: 4.2 MHz I signal: 1.5 MHz Q signal: 0.5 MHz.

In the invention of claim 4, only one threshold value needs to be provided, and the remaining I and Q threshold values are automatically determined. In particular It is. After that, the sampling can be controlled using the YIQ display system. The invention of claim 3 is this general concept.

In many cases, the Q component can be satisfied even without it, and this is the invention of claim 7, and the inventions of claims 5 to 6 are superordinate concepts. The reason why it can be made even without the Q component is described below. Let the colors of the two sample points be I1 and I2,
If the three primary color components are I1r, I1g, I1b, I2r, I2g, I2b, and the YIQ components are I1y, I1i, I1q, I2y, I2i, I2q, then (Equation 6) And the difference between the YIQ components is Becomes

When the difference between the three primary color components of the two sample points is represented by rgb (| I1r-I2r |, | I1g-I2g |, | I1b-I2b |), the maximum value of the difference between the YIQ components is Y Component: rgb (1,1,1) and 1 I component: rgb (1,0,0) or rgb (0,1,1) 0.596 Q component: rgb (1,0,1 ) Or rgb (0,1,0), which is 0.522. Applying the threshold determined by (Equation 8) to this, the difference of I component 0.596 is effective when the threshold of Y component = 0.596 / (4.2 / 1.5) = 0.213. The difference of I component, 0.596, is valid when the threshold of the Y component is 0.522 / (4.2 / 0.5) = 0.0621. Therefore, when the threshold of the Y component is 0.213 or less, the I component is significant, and when the threshold of the Y component is 0.0621 or less, the Q component is significant.

However, it is rare to give about 0.06 to the threshold value of the Y component, and about 0.1, and in many cases about 0.08 is sufficient. Given a small value like 0.06,
These small differences are usually present throughout the image,
Eventually, everything is sampled, and the gist of the present invention becomes meaningless. When 0.06 is given to the threshold value of the Y component, the threshold value of the I component is 0.168 and the threshold value of the Q component is 0.504 from (Equation 7). If there is a difference in the Q component even at 0.504, it is normal that the difference in the Y component is 0.06 or more and the difference in the I component is 0.168 or more, and the Q component has little meaning. In other words, it can be said that the Q component is unnecessary in a fine area such as anti-aliasing. Therefore, sampling control can be sufficiently achieved only with the Y and I components.

Incidentally, YIQ conversion is not used in HDTV, and the chromaticity equation of (Equation 10) is used. Hi-Vision chromaticity conversion formula Ey is the luminance signal component, Epr is the broadband chromaticity signal component, Epb
Is a narrow band chromaticity signal component. The bandwidth is enough if Ey signal: 20.0MHz Epr signal: 7.0MHz Epb signal: 5.5MHz. However, it is 30MHz in all studio standards. When the threshold is determined based on this Becomes

The colors of the two sample points are defined as I1 and I2.
If the primary color components are I1r, I1g, I1b, I2r, I2g, I2b, Ey, Epr, and Epb components are I1y, I1pr, I1pb, I2y, I2pr, I2pb, Becomes

When the difference between the three primary color components of two sample points is represented by rgb (| I1r-I2r |, | I1g-I2g |, | I1b-I2b |), the maximum difference between the Ey, Epr, and Epb components is obtained. The value is when Ey component: rgb (1,1,1) and 1 Epr component: when rgb (1,0,0) or rgb (0,1,1) 0.500 Epb component: rgb (0,0 , 1) or rgb (1,1,0), which is 0.500. When the threshold determined by (Equation 11) is applied to this, the difference of Epr component 0.500 is effective when the threshold of Ey component = 0.500 / (20/7) = 0.175. The difference of 0.500 between the Epb components is effective when the threshold value of the Ey component = 0.500 / (20 / 5.5) = 0.1375. Therefore, when the threshold value of the Ey component is 0.175 or less, the Epr component is significant, and the threshold value of the Ey component is 0.1375.
Epb components are significant when: Therefore, (Equation 10)
(Equation 12) can be used in the invention of claim 3,
It does not make much sense to apply this invention.

The function of the invention according to H1-2. The problem of edge recognition results in the problem of color difference. In the Japanese Industrial Standards JIS Z 8730-1980 color difference display method, the color difference expression using the L * a * b * display system, the color difference expression using the L * u * v * display system, the color difference expression of Adams-D.Nickerson, Hunter's (RSHunter) color difference formulas are used to determine the color difference, and sampling can be controlled by the magnitude of the difference. H1 ULCS (Unif
The orm Lightness Chromaticness Scale System) display system is a chromaticity diagram display system in which the chromaticity and the lightness are equi-rate. The display system of the chromaticity diagram with only equal chromaticity is UCS display system (Uniform Chromaticity Scale Sys
tem). ULCS display system is currently U * V * W * display system, L * a *
There are b * display system and L * u * v * display system, but they do not show sufficient perceptual difference. Humans can distinguish color difference from 1/5
It is said to be 1 / 10NBS unit.

[0037]

FIG. 3 shows an embodiment of the present invention.
The thresholds of R, G, and B are Tr, Tg, and Tb, respectively. S3-1
The threshold is set with, and this is adopted as it is as Tg.
In S3-2, Tr and Tb are obtained using (Equation 2). The determination of the colors I1 and I2 of the two sample points in S3-3 is one of actually calculating the luminance. The other uses the already calculated result. In S3-4, it is determined whether to perform oversampling. Simply performing subtraction may result in a negative value, so the absolute value is taken. The method of oversampling in S3-5 itself is outside the scope of the present invention. The scope of the present invention (claims 1 to 8 and H1 to 2) is over
It is a decision whether to sample or not. In S3-6, the above steps from S3-3 to S3-5 are repeated until the image generation is completed. In practice, oversampling may be nested, in which case it is determined whether to oversample each time.

FIG. 4 shows an embodiment of the third and fourth aspects of the present invention. The thresholds of the Y component, the I component, and the Q component are set to Ty, Ti, and Tq, respectively. The threshold is set in S4-1, and this is
To be adopted. In S4-2, Ty, Tq is obtained using (Equation 7). S4-3 is the same as S3-3. In S4-4, it is determined whether oversampling is performed using (Equation 9). S4
-5 is the same as S3-5, and S4-6 is the same as S3-6.

FIG. 1 shows an embodiment of the present invention. This is only because there is no element of Tq in FIG.
Is the same as

FIG. 5 shows an embodiment of the inventions H1-2. S5-3
Color difference formula in NBS unit, color difference formula in U * V * W * display system, color difference formula in L * a * b * display system, color difference formula in L * u * v * display system, Adams-Nickerson color difference formula, Apply one of the Hunter's color difference formulas. The rest is the same as the previous embodiment.

As described above, various methods of anti-aliasing have been described. In addition, there is a method that uses the difference between hue, lightness, and saturation, but it does not make much sense. Although the most accurate sampling control can be performed by using the color difference of the ULCS display system, the conversion from the RGB display system to the ULCS display system is troublesome, and the inventions of claims 4 and 7 are the most excellent.

Using one threshold T in the prior art, There is a technique for controlling oversampling by the following equation. This shows a behavior close to the case where the same threshold is given to all three in the technique of FIG. 2, but they are not necessarily the same. The fragile threshold must be given three times. In Figure 2
Although the three primary colors are determined independently, in this technique, a difference in one primary color may be compensated for by a difference in another primary color. Applying (Equation 1) to this There is also a method of controlling oversampling by the following equation. This means that only the Y component of the invention of claim 4 is used, which corresponds to the case where the threshold value of the Y component exceeds 0.213. Of course, anti-aliasing will be inadequate.

However, (Equation 9) does not completely match due to the relationship of the absolute values. Because | AB | + | CD | ≠ | A + CBD |, the meaning is different. For example, when A = 0, B = 10, | 0-B | + | C-0 | ≠ | 0 + CB-0 | | -B | + | C | ≠ | CB | B + C ≠ | CB | Because it becomes. B ≧ 0, C ≧
0, and B + C ≧ | CB |, that is, | AB | + | CD | ≧ | A + CBD |. Therefore, oversampling is performed extra as compared with the fourth aspect. That is, the efficiency does not increase at all. Therefore, care must be taken when practicing the invention of claims 4 and 7. It is important to use (Equation 9) as is and not to make a mistake in the order in which the absolute values are taken. This means that absolute values must not be taken when subtracting the three primary color components of two sample points.

The generally known CIE chromaticity diagram does not have uniform chromaticity and lightness. There is a problem because the NTSC YIQ display system was created based on this CIE chromaticity diagram.
For this reason, HDTV uses a different chromaticity equation.

The CIE was commissioned by the International Commission on Illumination (Commiss
FIG. 3 is a chromaticity diagram determined by Ion Internationale de l'Eclairage). Experiments with WDWright in 1941 and DLMacAdam in 1942 revealed the problem of CIE inequality. Middleton (WEKMiddletone) and Holmes (MCHo
lmes) in 1949. In 1954, an experimental NTSC color television broadcast was launched in the United States. The uv chromaticity diagram, one of the UCS chromaticity diagrams, was adopted by the International Commission on Illumination in 1960, and the L * a * b * display system, one of the ULCS display systems,
Recommended by the International Commission on Illumination in 1976. Hi-Vision experimental broadcasting was started in 1991, a historical background.

According to the experiment of 1949, the orange-cyan axis was determined to be highly discriminated and the yellow-magenta axis was discriminated to be low. From around 1976, the red-green axis (492-610 nm) was discriminated to be high and the yellow-blue axis was discriminated. (568
-465 nm) is considered low discrimination. This is the result of HDTV research.

In any case, anti-aliasing is not for the purpose of obtaining a complete color difference, but it is sufficient that there is no useless sampling. For this reason, the processing speed becomes slow, the calculation becomes complicated, and the program becomes difficult. It doesn't seem bloated and buggy.

Table 1 shows an example of the result of comparison between the anti-aliasing methods using the experimental method shown in FIG. The CPU used a 32-bit microprocessor called i486SX + i487SX (20MHz). The uniform random number is sran in Ref. [1].
The pseudo random number sequence generation method disclosed in the description of the d function was used. The random seed is 1. The method of generating real random numbers is disclosed in reference [2]. [1] American standard for information systems-programming language C, X3.159-1989 (America
nNational Standard for Information Systems − Prog
ramming Language C, X3.159-1989), pp.154-155 [2] Masahide Hirabayashi, "The Latest Program Dictionary in C Language, Volume 1"
Technical Review Co., Ltd., 1992, p.229-232

The given threshold is 0.1. The prior art is shown in FIG. 2, where the RGB thresholds are all 0.1 in this experiment. (Equation 13) was multiplied by 3 to give 0.3. S6-5 in Fig. 6 is S1-4, S2-3, S3-4, S4-4, S5-
Is replaced by the equivalent of 3, if this is true o
It increases s by one. os indicates the number of oversampling.

In the case of 640 × 400 pixels, the time related to the oversampling control takes only about 3 seconds at most as compared with the case where there is no control. Even if oversampling several times the number of pixels is not a problem. is there. Although the efficiencies are not simply comparable, if the appearances are assumed to be the same, the inventions of claims 4 and 7 are the most efficient. Similar results were obtained when the invention of claim 3 was implemented by (Equation 12). This experimental method reduces the brightness
Since the condition is to generate a uniform random number between 0 and 1,
The efficiency of real images will be much higher. In particular, yellow,
This is remarkable when generating images with many blues and purples.

This table points out that the time involved in the oversampling control is insignificant and almost out of the question, not to mention the efficiency. The effect of the present invention is not offset by the time required for oversampling control. In reality, the effect of anti-aliasing changes depending on the position of oversampling in an image, and the optimum value of the threshold ultimately becomes a matter of individual implementation. However, there is no change in that the effect of the present invention is sure.

[0052]

[Table 1]

[0053]

The effects of the inventions of claims 1-2. In the prior art, there is no use for blue, which has a low color difference discrimination ability for human eyes, because the same oversampling as green is performed. According to the invention of claim 2, blue is 5.15 times green and red is green
Oversampling is not performed unless there is a difference of 1.96 times. The present invention was able to reduce about 5% extra oversampling compared to the prior art. This is based on the experimental method of FIG. 6, and the efficiency of a real image is higher. This is particularly noticeable when generating an image with many yellows, blues, and purples. The present invention can also be used as a method for determining a threshold value of an image generation system already in operation.

The effects of the inventions of claims 3 to 7. In the prior art, similar oversampling was performed for RGB, despite the fact that the human eye's ability to discriminate color differences fell down and varied depending on chromaticity. In other words, over
As for items that cannot be distinguished even if they are sampled or not, there is no waste because of oversampling. Human eyes have the property that very fine things are judged only by brightness, relatively fine things are judged by the difference between green and blue, and even rougher things are judged by the difference between yellow and blue. In the present invention, oversampling is controlled in consideration of this, so that wasteful anti-aliasing can be realized. The present invention was able to reduce 16% extra oversampling compared to the prior art. This is due to the experimental method of FIG. 6, and the efficiency of a real image is higher, and a remarkable effect is exhibited when an image with many yellows, blues, and purples is generated.

Effects of the inventions of H1-2. The most accurate sampling control can be performed by using the color difference of the ULCS display system.

Although various anti-aliasing methods have been described in this specification, the most superior ones are the inventions of claims 4 and 7. The present invention is simple to implement, has a certain effect, and is applicable to all anti-aliasing methods based on oversampling.

[Brief description of the drawings]

FIG. 1 is a flowchart showing an embodiment of the invention according to claims 5 to 7;

FIG. 2 is a flowchart showing a conventional technique according to the first and second aspects of the present invention;

FIG. 3 is a flowchart showing an embodiment of the invention according to claims 1 and 2;

FIG. 4 is a flowchart showing an embodiment of the invention according to claims 3 and 4;

FIG. 5 is a flowchart showing an embodiment of the invention H1-2.

FIG. 6 is a flowchart showing an experimental method of the present invention.

Claims (7)

(57) [Claims] In anti-aliasing by oversampling, red, green, and blue values at two sample points are separately compared, and any one of the differences is calculated.
In an anti-aliasing method that performs oversampling when one exceeds (or exceeds) a preset threshold, a reference green threshold is set, and a red and blue threshold is a green threshold. An anti-aliasing method characterized in that weighting is performed according to the color difference discrimination ability of the human eye based on the threshold value, and the red threshold value is larger than the green threshold value, and the blue threshold value is larger than the red threshold value. 2. The anti-aliasing method according to claim 1, wherein the red threshold is a green threshold 0.587 / 0.299, and the blue threshold is a green threshold 0.587 / 0.114. Anti-aliasing method. 3. An anti-aliasing method based on oversampling, wherein a luminance component indicating the brightness of two sample points, a chromaticity component having high human eye discrimination ability, and a low human eye discrimination ability are provided. The chrominance components are separated into three components, and each of the three components is compared separately. If any one of the differences exceeds (or exceeds) a preset threshold, oversampling is performed. An anti-aliasing method, comprising: However, the three thresholds set in advance are increased in the order of a luminance component indicating brightness, a chromaticity component having high human eye discrimination ability, and a chromaticity component having low human eye discrimination ability. 4. An anti-aliasing method based on oversampling, wherein one threshold is set, the threshold is set as a Y component threshold, and the I component threshold is set as the Y component threshold / Y signal bandwidth / I signal bandwidth. Bandwidth, Q
The threshold of the component is determined as the threshold of the Y component, the bandwidth of the Y signal / the bandwidth of the Q signal, and three thresholds are determined. The colors of the two sample points are converted from the RGB display system to the YIQ display system. An anti-aliasing method comprising comparing each of the components separately and oversampling if any one of the differences exceeds (or exceeds) a respective threshold. 5. An anti-aliasing method based on oversampling, wherein a chromaticity component having a low human eye discrimination ability is ignored from colors of two sample points, a luminance component indicating brightness, a human eye discrimination. Take out two of the chromatic components with high power, compare each of the two components separately, and if any one of the differences exceeds (or exceed) the preset thresholds, An anti-aliasing method characterized by sampling. 6. The anti-aliasing method according to claim 5, wherein a threshold value of a luminance component indicating brightness is smaller than a threshold value of a chromaticity component having high human eye discrimination ability. Anti-aliasing method. 7. An anti-aliasing method based on oversampling, wherein one threshold is set, this is set as a threshold of a Y component, and a threshold of an I component is set as a threshold of a Y component · 4.2 / 1.5 to determine two thresholds. Then, Y and I of the YIQ display system are extracted from the colors of the two sample points, Y and I components are separately compared, and one of the differences between Y and I is obtained.
An anti-aliasing method characterized in that oversampling is performed when each threshold value is exceeded (or is equal to or greater than the threshold value).