KR100597156B1 - Method and system for improving the display resolution of an image using subpixel sampling and visible error filtering - Google Patents

Abstract

The present invention provides a method and apparatus for converting a high resolution color image into a low resolution image. According to the method of the present invention, based on the characteristics of the spatial frequency response of the human visual system, the low resolution image 116 is appropriately sampled and filtered for each of the luminance channel 76 and the chrominance channel 98, 100. Formed by separating the high resolution image 72 into the luminance channel 76 and the chrominance channels 98 and 100, and combining the luminance channel 86 and the chrominance channels 98 and 100 after sampling. do. Typical subsampling 101,103 is performed on chrominance channels 98 and 100 to avoid chromatic aliasing, while subchannel sampling 80 is performed on luminance channel 76 to improve the resolution of the luminance component. ).

Resolution Conversion, Chrominance Channel, Luminance Channel, Color Domain

Description

METHODS AND SYSTEMS FOR IMPROVING DISPLAY RESOLUTION IN IMAGES USING SUB-PIXEL SAMPLING AND VISUAL ERROR FILTERING}

The present invention displays a high resolution image, which uses a triad arrangement to display R, G, B or other components of the image. This triad arrangement is much of what is seen, for example, in direct viewing LCD displays, where a single pixel consists of three sub-pixels located side by side. Each subpixel controls only one of the three primary colors (i.e., R, G, B), i.e. each subpixel is generally controlled only by the three primary colors of the digital image representation. High resolution images are available in memory, or directly from algorithms (vector graphics, some font designs, and computer graphics).

The title of the present application is related to US Patent Application No. 09 / 735,454 filed on December 12, 2000, entitled "Methods and Systems for Improving Display Resolution using Sub-Pixel Sampling and Visual Error Compensation" by Scott Daly. This application is incorporated herein by reference.

Also, the title of the present application is US Patent Application No. 09 / 735,425 entitled "Methods and Systems for Improving Display Resolution in achromatic Images using Sub-Pixel Sampling and Visual Error Filtering" by Rajesh Reddy K Kovvuri and Scott Daly. 2000.12.12). This application is incorporated herein by reference.

The most commonly used method for displaying a high resolution image on a low resolution display is to sample the pixels 2 of the high resolution image 4 as shown in FIG. The R, G, and B values of each downsampled color pixel 8 are then mapped to the individual R, G, B elements 10, 12, 14 of each display pixel 16. Since the display device does not allow overlapping of color elements, the subpixels can only take one of three R, G, and B colors, but the amplitude of the color may change over the entire grayscale range (i.e. 0 to 255). Can be. The subpixels typically have an aspect ratio of 1: 3 (width: height), so that the resulting pixel 16 is square. The subsampling / mapping technique does not take into account that the R, G, and B subpixels of the display are spaced apart, and in fact is assumed to overlap in the same way as in a high resolution image as shown in FIG. This kind of sampling is called subsampling or traditional subsampling.

The pixels of the high resolution image 4 are shown as three slightly offset stacked squares 8 to indicate that their RGB values are related for the same spatial location (ie pixels). One display pixel 16 composed of R, G, and B subpixels 10, 12, 14 is shown as part of the low resolution triad display 6 using thick lines in FIG.

In this example, the high resolution image has 3x or more resolution than the display (in horizontal and vertical dimensions). Since this direct subsampling technique causes aliasing artifacts, various methods of averaging neighboring unsampled pixels with sampled pixels are used. Note that the conventional technique of averaging neighboring elements during subsampling is mathematically equivalent to prefiltering a high resolution image with a rectangular filter. It should also be noted that the technique of selecting a pixel other than the leftmost (as shown in FIG. 1) can be considered as prefiltering that only affects phase. Thus, much of the processing associated with preventing aliasing can be considered as a filtering operation for high resolution images, even though the kernel is only applied at the sampled pixel positions.

Achromatic images as defined herein do not have visible color changes. This achromatic condition may occur when the image contains only one layer or color channel, or when the image includes multiple layers or multiple color channels, but each color layer is the same to produce a single color.

It has been found that the above technique does not exploit the potential display resolution. Background information in this area can be found in R. Fiegenblatt's "Full color imaging on amplitude color mosaic displays" Proc. SPIE V. 1075, 199-205 (1989), and J. Kranz and L. Silverstein, "Color matrix display image quality: The effects of luminance and spatial sampling" SID Symp. See Digest 29-32 (1990), which is incorporated herein by reference. For example, in the display shown in FIG. 1, when the resolution of the display pixel 16 is 1/3 of the high resolution image (source image) 4, the subpixels 10, 12, 14 are (horizontal dimensions). Has the same resolution as that of the source). If this display is used only by individuals who are color blind, it would be possible to use the spatial location of that subpixel. This scheme is shown in FIG. 2, where the R, G, B subpixels 10, 12, 14 of the display are taken from the corresponding colors of the different pixels 11, 13, 15 of the high resolution image. For this reason, the horizontal resolution becomes subpixel resolution, which is 3x the display pixel resolution.

But what about display observers who are not color blind? That is, most observers. Fortunately for display engineers, even observers with full color 5 vision are color blind at the highest spatial frequencies. This is shown below in FIG. 3, which shows the ideal spatial frequency response of the human visual system.

Here, the luminance 17 represents the achromatic contact of the observed image, and the chrominance 19 represents the color content, which is visualized by the visual system from "red" to "green" and from "blue" to "yellow". It is treated like isoluminant modulation. The color difference signals (R-G, B-Y) of the video are approximate to this modulation. For most observers, the bandwidth of the chromatic frequency response is half the bandwidth of the luminance frequency response. Sometimes the bandwidth of the blue-yellow modulation response is much smaller, down to about one third of the luminance. As shown in Fig. 2, the sampling including the mapping of color elements from different image pixels to subpixels of the display pixel triad may be referred to as "subpixel sampling".

Referring to FIG. 4, in the horizontal direction of the display, the Nyquist of the display pixel 16 (display pixel = triad pixel, wherein the triad Nyquist is 0.5 cycles per triad pixel) and the subpixel elements 10, 12. There is a frequency range between the Nyquist frequency (0.5 cycles per subpixel = 1.5 cycles / tried pixels). This area is shown as blind area 20 in FIG. 4. The sinusoidal function arising from convolving a high resolution image into a rectangular function of the same width as the display sample interval is shown by dashed line 22. This is the most common way to model the display MIF (Modulation Transfer Function) when the display is an LCD.

The sinusoidal function resulting from convolving a high resolution source picture into a quadrature function equal to the subpixel spacing is shown by dashed line 24 and has a high bandwidth. This is a limitation imposed by the display considering that the subpixels are at right angles in one dimension. In the rectangular area 20 shown, the subpixels can display luminance information, but cannot display chrominance information. Therefore, by performing chrominance aliasing in this region, it is possible to obtain frequency chrominance information that is allowed by triad (i.e., display) pixels. This is an advantage area obtained by using subpixel sampling.

In applications with font displays, typically black & white fonts are preprocessed as shown in FIG. Standard preprocessing is a hint that centers the font stroke at the center of the pixel, i.e., a font-stroke specific phase shift. Subsequently, low pass filtering is generally performed, which is called grayscale antialiasing processing.

The visual frequency responses (CSFs) shown in FIG. 3 are ideal. In practice, the response has a finite falloff slope as shown in Fig. 6 (a). The luminance CSF 30 is mapped from the cy / deg unit to the display pixel domain (assuming an observation distance of 1280 pixels). This is shown by the solid line 30 with the maximum frequency near 1.5 cy / pixel (display pixel) and bandpass in the form with a peak near the 0.2 cy / pixel triad. R: G CSF 32 is shown in dashed lines and lowpasses at a maximum frequency near 0.5cy / pixel. The B: Y modulated CSF 34 is shown as a dashed-dot LPF at a maximum frequency similar to the R: G CSF, but with a lower maximum response. The range between the cutoff frequencies of the chrominance CSF 32, 34 and the luminance CSF 30 is a range that can allow chromatic aliasing to improve the luminance bandwidth.

6 (a) also shows the ideal image power spectrum 36 as a 1 / f function, which is shown as a straight line with a slope of −1 (since the logarithmic coordinate axis is used). This spectrum will be repeated at the sampling frequency. This repetition is shown for the sampling rates of subpixel 40 and pixel 38 in the horizontal direction. Repetitions occurring at low frequency 38 are due to pixel sampling and repetitions occurring at high frequency 40 are due to subpixel sampling. Note that the shape changes because it is plotted on the logarithmic frequency axis. The frequencies of this repetitive spectrum that extend down to the low frequencies below Nyquist are called aliasing. The leftmost one is due to the pixel sampling rate, which is chromatic aliasing 38, and the luminance aliasing 40 occurs at high frequencies because it is related to the high sub-pixel sampling rate.

In FIG. 6A, no prefiltering is applied to the source spectrum. As a result, aliasing due to pixel sampling (ie, chromatic aliasing) extends to a very low frequency 35. Thus, even if the chromatic CSF has a lower bandwidth than the luminance CSF, the color artifacts will still be visible (depending on the noise and contrast of the display).

In FIG. 6B, a prefilter (shown as the diagonal line 22 in FIG. 4) (the same orthogonal function as three source picture pixels) was applied to the source power spectrum, affecting the band spectrum 42 above 0.5 cy / pixel. It can be seen that, as indicated by the sign 44 has a steeper slope than -1. This repetition also shows the effect of the prefilter. Even with this filter, it can be seen that some chromatic aliasing (repeated spectrum at low frequencies) may occur at the cutoff frequency low frequencies 46 of the two chromatic CSFs 32a and 34a. Thus, it can be seen that simple luminance prefiltering is difficult to eliminate chromatic aliasing without removing all luminance frequencies above 0.5 cy / pixel (ie, the benefit region described above).

Since it depends on the visual system difference of the bandwidth as a luminance or chromatic function to provide an increase in the luminance bandwidth in the benefit area 20, the contents of which are incorporated herein by reference and are described in C. Betrisey et al. Displaced filtering for patterned displays There is a possibility to design prefiltering based on a visual system model as described in SID Symposium Digest, 296-299 (2000).

This technique ideally uses different prefilters depending on the color layer and the color subpixel from which the image is being sampled. Thus, there are nine filters. These are described by X. Zhang and B. Wandell, "A spatial extension of CIELab for digital color image reproduction", SID Symposium Digest 731-734 (1996), the contents of which are incorporated herein by reference and incorporated by reference. It was designed using the human visual difference model described in. This was done offline and assumes that the picture is always black & white. In the final implementation, orthogonal functions are used to reduce the computation rather than the resulting filter. In addition, there is still some residual chromatic error that can be seen as chromatic aliasing extends to the chromatic CSF cutoff frequency (as can be seen in FIG. 6B). However, the visual model used did not take into account the masking properties of the visual system, which caused masking of chrominance by luminance when the luminance was in the middle of the high contrast level. Thus, larger font chromatic artifacts present along the edge of the font are masked by the high luminance contrast of that font. However, as the font size decreases, the luminance of the font also decreases, so that the same chromatic artifacts look very good (eg, in very small fonts). (The black / white portion of the font disappears, leaving only local small color spots.)

The present invention has been invented on the basis of the above characteristics of the spatial frequency response of the human visual system, namely the luminance CSF has a higher cutoff frequency than the chrominance CSF. In the present invention, in order to perform appropriate sampling and filtering on each of the luminance data and chrominance data, the low resolution image is divided into a high resolution image into luminance data and chrominance data, and after sampling the luminance data and chrominance data. It is formed by combining. For chrominance data, traditional subsampling is performed to avoid chromatic aliasing, while subpixel sampling is performed on the luminance data to improve the resolution of the luminance component. In addition, high-pass filtering is performed on the sampled subpixel luminance data so as to remove low frequency artifacts generated during subpixel sampling of the luminance data.

The concept of the present invention is explained below. 1 and 2, each of the R, G, and B values of the R, G, and B subpixels 10, 12, and 14 in one display pixel 16 has a high resolution by subsampling shown in FIG. R, G, and B values of the color pixels 8 (that is, 11) of the image 4 having? However, each of the R, G, and B values of the subpixels 10, 12, and 14 is not equal to the R, G, and B values of the color pixels 8 (11). Therefore, in terms of the luminance component, the subpixel sampling shown in FIG. 2 is performed so that each of the R, G, and B values of the subpixels 10, 12, and 14 corresponds to the luminance component of the color pixels 11, 13, and 15. Reflect.

Embodiments of the present invention include methods and systems that are less dependent on linearity assumptions and filtering and can operate on input color pictures. This embodiment can directly remove low frequency chromatic artifacts caused by subpixel sampling. This is possible by creating an LPF version of the chromatic content of the picture that is added to the luminance and chromatic aliasing versions. This is done by using color dominance, i.e., primary color domain (i.e., RGB), rather than false color, to remove color artifacts caused by subpixel sampling. In fact, only the low frequency chromatic artifacts need to be deleted, as high frequency chromatic artifacts are not visible due to the low bandwidth of the chrominance CSF as shown in FIG. 6A.

The method and system of the present invention can be used to obtain a high resolution luminance signal in which no chromatic aliasing is seen when the display is not observed closer than the design specification. This technique need not assume that the source picture is text or the picture is achromatic.

Embodiments of the present invention convert high resolution pictures into low resolution pictures with reduced errors caused by the subsampling process. If the high resolution image is not in a format capable of separating luminance and chrominance data, the image is converted to such a format. Many complementary domains are acceptable. The complementary color domain image is divided so that separate processing is possible by separating the luminance channel from the chrominance channel.

The luminance channel is then converted to a false color domain (ACD) such as RGB, and the ACD luminance image is subpixel sampled to maintain the luminance data while lowering the resolution. After subpixel sampling, the subpixel sampled (SPS) image is converted back into the complementary color domain (OCD) and divided into separate luminance and chrominance channels. The SPS chrominance channel caused by this division is highpass filtered to remove low frequency artifacts caused by subpixel sampling. The SPS luminance channel typically remains unchanged and retains the original luminance data.

The chrominance channel from the original picture is lowpass filtered and then subsampled to produce chrominance data for the low resolution picture. This lowpass filtered chrominance channel is combined with a highpass filtered subpixel sampled chrominance channel, originally generated from the luminance channel. This combined chrominance channel is also combined with the SPS luminance channel to form a low resolution image with generally reduced error in the complementary domain. The reduced resolution image of this reduced error is then converted into a false color domain or some other color domain compatible with the desired application.

BRIEF DESCRIPTION OF DRAWINGS To obtain the above and other advantages and objects of the present invention, a more detailed description of the present invention will be described with reference to specific embodiments illustrated in the accompanying drawings. These drawings illustrate only some embodiments of the invention and do not limit the scope of the invention, which is described with reference to the accompanying drawings.

1 illustrates traditional image sampling for a display having a triad pixel configuration.

2 shows subpixel image sampling for a display having a triad pixel configuration.

3 is a graph showing an ideal CSF mapped to the digital frequency domain.

4 is a graph showing region analysis of pixel Nyquist and subpixel Nyquist representing benefit regions.

5 illustrates a conventional pretreatment technique.

FIG. 6A is a graph showing analysis using 1 / f power spectrum repeated at pixel sampling and subpixel sampling frequencies. FIG.

FIG. 6B is a graph showing analysis using 1 / f power spectrum repeated at subpixel sampling frequency and pixel sampling due to improved preprocessing. FIG.

7 is a block diagram of using a known visual model.

8 is a flow diagram illustrating an embodiment of the invention.

9 is a flow diagram illustrating a particular embodiment of the present invention.

10 is a graph showing a signal maintained by an embodiment of the present invention.

Preferred embodiments of the present invention will be understood with reference to the drawings, wherein like parts are designated by like numerals throughout the drawings. Each of the above figures is included as part of this specification.

In general, as shown in the drawings, the components of the present invention may be variously modified. Accordingly, the following description of embodiments of the methods and systems of the present invention is not intended to limit the present invention, but is merely an example of a preferred embodiment of the present invention.

Embodiment elements of the invention may be embodied as hardware, firmware and / or software. In addition, the software (program) on which the features of the embodiments of the present invention are implemented can be stored and provided in a computer-readable recording medium. Examples of such media include computer media (LAN, WAN, and wireless as well as recording media such as information storage means (semiconductor memory, floppy disks, hard disks, etc.) and optical storage means (CD-ROM, DVD, etc.). Communication network) communication media (optical fibers, wireless communication lines, etc.) used in the system. In addition, features of embodiments of the present invention can be implemented as computer signals embodied in electronic transmission. The exemplary embodiments described herein have merely described one of such forms and those skilled in the art will understand that such elements may be achieved by any of these forms within the scope of the present invention.

Embodiments of the present invention are described and claimed with respect to the term achromatic. As used in connection with an image in the present specification and claims, this term refers to an image without visible color variation. An achromatic image may be an image comprising only one layer or color channel, or an image comprising multiple layers or color channels, where each color layer is the same and thus produces a single color.

Embodiments of the present invention are described and claimed in the context of "RGB" pictures or domains, or "color domains," or "color gamut pictures." These terms, as used herein and in the claims, may refer to any form of a multi-component picture domain with integrated luminance and chrominance information, including but not limited to various RGB domains and CMYK domains. .

Embodiments of the invention are also described and claimed in the context of "YCrCb" pictures or domains, "Complementary" domains, pictures or channels, or "color difference" domains, pictures or channels. As used herein and in the claims, these terms include a multi-component picture domain having a channel comprising individual luminance channels and at least one chrominance channel, including but not limited to YCrCb, LAB, YUV, and YIQ. It can point to any form of.

Embodiments of the present invention can be used to convert a high resolution image into a low resolution image with less visible error in the converted image. These embodiments are typically used in connection with a display device to convert an image having a display high resolution to lower to a resolution usable by the display, but other applications are possible.

The converted image according to the embodiment of the present invention may exist in various formats. When such a format is not compatible with the process of the embodiment of the present invention, the image can be converted to a compatible format before being processed, and then converted again after processing as necessary.

An embodiment of the present invention will be described with reference to FIG. 8 showing an overview of the embodiment. Process 70 begins with an image present in a complementary domain (OCD) or similar domain, such as YCrCb, LAB, YUV, YIQ. When an image is in a false color domain (ACD), such as an RGB or CMYK domain or some other color space, the image may be converted to a complementary color domain before being processed according to an embodiment of the present invention. Some embodiments include converting the image into a compatible format prior to processing.

Once the picture is in the complementary domain 72 having separate luminance and chrominance channels, the picture is divided into a luminance channel and a chrominance channel by the separation process of step 74. The division process (step 74) may include sampling or filtering the original OCD picture 72 or separating the luminance and chrominance data from the original picture 72. In addition, the division process may include image conversion.

After the division process, the initial luminance channel 76 is converted into an ACD luminance image such as an RGB image (step 78). This enables sampling of the luminance picture in the domain or format that is finally displayed. Once the luminance image is converted (step 78), the resulting low resolution image is subjected to subpixel sampling on the image to improve the resolution (step 80). In this way, the luminance data from each successive pixel in the high resolution original image is assigned to each corresponding subpixel in the low resolution image.

When subpixel sampling (step 80) is completed, a subpixel sampled (SPS) luminance image is obtained as a result, and this SPS luminance image is converted into an OCD image, which can be referred to as an SPS-OCD luminance image (step 82). ). When this conversion is performed, the SPS luminance picture may be further divided into separate luminance channels and chrominance channels (step 84). The SPS luminance channel 86 typically remains intact until subsequent coupling with other channels (step 88). However, the SPS chrominance channels 90 and 92 are filtered before further combining.

These SPS chrominance channels 90, 92 can be separated into red-to-green channels 90 and blue-to-yellow channels 92. . These channels are typically Cr and Cb channels of YCrCb pictures, "a" and "b" channels of LAB pictures, U and V channels of YUV pictures, I and Q channels of YIQ pictures, or similar in other color spaces or domains. Include channels. These chrominance channels 90, 92 are high pass filtered (steps 94, 96) to remove low frequency artifacts that occur during subpixel sampling.

In some embodiments of the present invention, high pass filtering (steps 94 and 96) may be performed by an unsharp mask method. Unsharp masks can use a lowpass kernel. Typically, the original picture is processed with a lowpass kernel to produce a lowpass version of the picture. Subsequently, this lowpass version is removed from the original unfiltered picture while maintaining the average value of that picture. In a successful embodiment, a Gaussian lowpass kernel with a sigma of about 0.3 pixels to about 0.8 pixels was used. In particular, a sigma value of 0.6 pixels is considered successful, resulting in a cutoff in the frequency domain of about 0.168 cycles / pixel. This provides a good unsharp mask filter. Deploying the Gaussian kernel looks like this:

The one-dimensional Gaussian function used in some embodiments is as follows.

Fourier transform this function:                 

Here, it can be seen that σ in the spatial domain (pixel unit) corresponds to 1 / π ² σ in the frequency domain (pixel / cycle unit). This relationship can be used to determine the cutoff frequency of the filter in the case of sigma, or vice versa for the unsharp mask that can be derived by the CSF model in the case of frequency.

The two-dimensional Gaussian function used in some embodiments is as follows.

Since this Gaussian function is a separable Cartesian function, the frequency response of the two-dimensional Gaussian function is similar to Equation 2 given the meaning of σ. That is, σ _x in the time domain is 1 / π ² σ _x in the frequency domain, and σ _y in the time domain is 1 / π ² σ _y in the frequency domain.

In a successful embodiment of the present invention, a Gaussian unsharp mask filter implemented with a 3x3 kernel and having a sigma value of 0.6 was employed to obtain a cutoff frequency of a lowpass filter of about 0.2 cycles / pixel.

Another embodiment of the invention may use a highpass filter that is equivalent to the reverse CSFs for each of the complementary channels. These CSFs can be mapped from the domain of cy / deg (self modeled) to the digital domain of cy / pix. The actual mapping process takes into account the viewing distance and can be individualized for different applications, with specific display resolutions in pixels / mm and different expected or intended viewing distances. As a result of the method of the invention, the chromatic artifact will not be visible when not observed at a distance closer than the design observation distance. However, the luminance resolution will be improved.

This filtering process (steps 94, 96) may be performed for all chrominance channels 90, 92 or selected channels based on the amount or intensity of artifacts generated in a particular sampling process, or some other criterion.

The low pass filtering process (steps 102, 104) of the original OCD chrominance channel 98, 100 may be performed concurrently with the luminance path 105 process or may be performed at different times. Lowpass filtering processing (steps 102 and 104) of the OCD chrominance channel is performed to remove significant chromatic frequencies above the display pixel Nyquist frequency. Thus, these channels may be subsampled by a 1: 3 factor in the traditional manner without generating chromatic aliasing in the chromatic path 110 (steps 101 and 103).

Once the filtering operation is completed, the separated channels can be combined. The combination of chrominance channels will change depending on the color domain used. In this embodiment, the highpass filtered and subpixel sampled blue-to-yellow (HPFSPS-B / Y) chrominance channel 97 is a lowpass filtered and traditionally subsampled blue-to-yellow ( LPFSS-B / Y) is combined with chrominance channel 109 (step 106) to form a single high pass filtered (HLF) B / Y chrominance channel 111. In addition, the highpass filtered and subpixel sampled red-to-green (HPFSPS-R / G) channel 95 is a lowpass filtered and traditionally subsampled red-to-green (LPFSS-R / G) channel. 107 is combined (step 108) to form a single high pass-low filtered (HLF) R / G channel 114.

It should be noted that the method of embodiments of the present invention may be used in other color spaces and domains, which may include other color channels and different amounts of color channels as well as variations in luminance or brightness channels. The combined HLF chrominance channels 111, 114 may be further combined with the SPS luminance channel 86 to form a low resolution OCD picture 116 (step 88). The low resolution OCD picture 116 can then be converted or converted to other picture formats or domains for various purposes.

The method and system of these embodiments provide a low resolution image with fewer visible chromatic artifacts.

In addition, the above-described embodiments may be changed in various ways. For example, in some embodiments, low pass filtering (steps 102, 104) of the chrominance channels 98, 100 may be omitted. Further, by using only the luminance channel 76 separated in step 84, a low resolution image 116 can be formed. That is, combining the SPS luminance channel 86, the HPFSPS-R / G channel 95 and the HPFSPS-B / Y channel 97 while performing each step of the luminance path 105 with respect to the luminance channel 76. To do this, steps 106 and 108 may be omitted for each step of the chrominance path 110 and for the chrominance channels 98 and 100. In this way, a low resolution image 116 can be formed.                 

9, a specific embodiment of the present invention will be described. This embodiment can be used to process high resolution RGB images for display on low resolution display devices. The high resolution RGB image 120 may optionally be preprocessed according to the specific needs of the user or application (step 122). Preprocessing (step 122) includes hints, types of lowpass filtering, or other processing techniques. In addition, the preprocessing (step 122) may be bypassed completely.

After preprocessing (step 122), RGB may be converted to a complementary color domain picture such as LAB, YCrCb, YIQ, YUV or other picture domains (step 124). In this example, the LAB picture domain is used. Once switched to this domain, the picture is divided into individual L, a, b channels of this domain and subjected to individual processing of each channel (step 126). In this way, chrominance and luminance channels can be processed separately.

The L channel 127 is then switched back to the RGB domain so that it can be sampled to the final display format. This conversion involves simply copying the L layer or channel into three identical R, G, and B layers. In addition, although a single layer can be used, the actual conversion method depends on the selected color conversion.

Then, a subpixel sampling process (step 130) is performed on the RGB luminance image to maintain the horizontal luminance resolution of the original RGB image 120. After the subpixel sampling process, the sampled image is switched back to the complementary color domain such as LAB (step 132). The sampled LAB picture is divided (step 134) to separate the luminance and chrominance channels for further processing of the chrominance channel. Here, the luminance channel 136 is typically not processed and retains the original luminance data. However, the luminance channels 150, 152 of the subpixel sampled and divided images are high pass filtered (steps 146, 148) to remove low frequency chromatic aliasing that occurs during subpixel sampling.

As in the other embodiments described above, this highpass filtering process can be performed with an unsharp mask filter using a Gaussian lowpass kernel. In an embodiment using this method, the chrominance channel is filtered to filter the lowpass filtered from the SPS-RGB chrominance picture to produce a highpass filtered (HPF) SPS chrominance picture or channels 147 and 149. Generate a chrominance image. The highpass filtering process (steps 146, 148) is typically performed for the "a" and "b" channels, but can only be performed for one channel if conditions permit.

The low pass filtering process (steps 138, 140) of the original "a " and " b " chrominance channels 154, 156 may be performed simultaneously with the processing of the " L " channel or at different time zones. Lowpass filtering processing (steps 138, 140) of the " a "and " b " chrominance channels is performed to remove significant chrominance channels above the display pixel Nyquist frequency. After the lowpass filtering process (steps 138 and 140), these channels can be subsampled by a 1: 3 factor without the occurrence of chromatic aliasing in the traditional manner (steps 142 and 144).

Once the channels are filtered and sampled, they are combined to form a low resolution picture with less error. The highpass filtered luminance "a" channel 147 is subsampled and combined with the lowpass filtered "a" channel 143 (step 160) to form a processed "a" channel 164. The highpass filtered luminance "b" channel 149 is combined with the subsampled and lowpass filtered "b" channel 145 (step 158) to form a processed "b" channel 162. These chrominance channels 162 and 164 then combine with the SPS luminance channel 136 (step 166) to form a low resolution LAB image 168 with reduced error.

This error reduced image can be switched to the RGB domain (step 170) to produce an error reduced low resolution RGB image 172 that can be output to a display or other device.

The process function of an embodiment of the present invention is described with reference to FIG. 10, which shows the signal maintained for the luminance CSF 180 and the chrominance CSF 182. As shown in FIG. The retained chromatic signal includes a lowpass region 186 containing useful chromatic content of the image, as well as a highpass region 184 that is not detected for chrominance CSF. Ideally, frequencies deviating from this lowpass region 186 will not be visible to the observer because of the limited bandwidth of the RG and BY CSF. The HPF chromatic signal 184 is chromatic aliasing that conveys valid luminance information. Note that this technique works well for color pictures, since low frequency chromatic information is maintained. Although Fig. 10 does not show the overlap between these two chromatic signals, the overlap may occur depending on the filter actually used. In other embodiments, a filter may be used that allows for overlap of the chromatic signals of the highpass region 184 and lowpass region 186 shown in FIG. By overlap, the chromatic bandwidth can be widened even if the occurrence of the chromatic aliasing is taken into account.                 

The present invention can be embodied in other specific forms without departing from the spirit or essential features of the invention. The above embodiments are merely exemplary in all respects and the present invention is not limited thereto. Accordingly, the scope of the invention is limited by the claims rather than the foregoing description. All changes that come within the meaning and range of equivalency of the claims are to be embraced within their scope.

Claims (20)

A method of converting a high resolution image into a low resolution image with reduced visual error, Separately dividing a high resolution complementary domain (OCD) image into one initial luminance channel and one or more initial chrominance channels; Generating an additive color domain (ACD) luminance image by performing subpixel sampling on the initial chrominance channel; Converting the ACD luminance picture to an OCD luminance picture; Separately dividing the OCD luminance picture into one subpixel sampled (SPS) luminance channel and one or more SPS chrominance channels; Performing high pass filtering on the one or more SPS chrominance channels; And Combining the high-pass filtered one or more SPS chrominance channels with the SPS luminance channel to form a low resolution image with reduced error Method comprising a. An apparatus for converting a high resolution image into a low resolution image having a reduced visual error, Means for separately dividing a high resolution complementary domain (OCD) picture into one initial luminance channel and one or more initial chrominance channels; Means for performing subpixel sampling on the initial luminance channel to generate an additive color domain (ACD) luminance image; Means for converting the ACD luminance picture to an OCD luminance picture; Means for separately dividing the OCD luminance picture into one subpixel sampled (SPS) luminance channel and one or more SPS chrominance channels; A high pass filter for performing high pass filtering on the one or more SPS chrominance channels; And Means for combining the high-pass filtered one or more SPS chrominance channels with the SPS luminance channel to form a low resolution image with reduced error Apparatus comprising a. A computer readable medium storing a method for converting a high resolution image into a low resolution image with reduced visual error, the method comprising: Separately dividing a high resolution complementary domain (OCD) picture into one initial luminance channel and one or more initial chrominance channels; Generating sub-color domain (ACD) luminance images by performing subpixel sampling on the initial luminance channel; Converting the ACD luminance picture to an OCD luminance picture; Separately dividing the OCD luminance picture into one subpixel sampled (SPS) luminance channel and one or more SPS chrominance channels; Performing high pass filtering on the one or more SPS chrominance channels; And Combining the highpass filtered one or more SPS chrominance channels with the SPS luminance channel to form an error reduced low resolution image Computer-readable medium comprising a. A method of converting a high resolution image into a low resolution image with reduced visual error, Dividing a high resolution complementary domain (OCD) picture into separate initial luminance and initial chrominance channels; Generating a complementary color domain (ACD) luminance image by performing subpixel sampling on the initial luminance channel; Converting the ACD luminance picture to an OCD luminance picture; Dividing the OCD luminance picture into individual subpixel sampled (SPS) luminance and SPS chrominance channels; Performing high pass filtering on the SPS chrominance channel; Performing lowpass filtering on the initial chrominance channel; Performing subsampling on the filtered initial chrominance channel; Combining the subsampled lowpass filtered chrominance channel with the highpass filtered SPS chrominance channel; And Combining the combined chrominance channel with the SPS luminance channel to form a low resolution image with reduced error Method comprising a. The method of claim 4, wherein And said high resolution image is a false color domain image that is converted to a complementary color domain image prior to said step of dividing a high resolution OCD image. The method of claim 4, wherein And said false color domain picture is an RGB picture. The method of claim 4, wherein The complementary color domain image is a YCrCb image. The method of claim 4, wherein And the complementary color domain picture is a LAB picture. The method of claim 4, wherein And said highpass filtering comprises unsharp mask filtering. The method of claim 4, wherein The high pass filtering step, Filtering the SPS chrominance channel through an unsharp mask filter having a Gaussian lowpass kernel to generate a lowpass chrominance channel; And Extracting the SPS lowpass chrominance channel from the SPS chrominance channel to produce a highpass filtered SPS chrominance channel. The method of claim 4, wherein And the chrominance channel comprises a red-green channel and a blue-yellow channel. The method of claim 4, wherein Wherein the chrominance channel comprises Cr and Cb channels of a YCrCb picture. The method of claim 4, wherein Wherein said chrominance channel comprises "a" and "b" channels of a CIELab picture. The method of claim 4, wherein Converting the error reduced low resolution image to an RGB image. The method of claim 4, wherein The performing of subpixel sampling may include: Converting the initial luminance channel into a false color domain (ACD) luminance picture and sampling the ACD luminance picture. A method of displaying a high resolution image at low resolution with reduced visual error, Converting a high resolution RGB image into a high resolution complementary domain (OCD) image; Dividing the high resolution OCD picture into separate initial luminance and initial chrominance channels; Converting the initial luminance channel into an RGB luminance picture; Performing subpixel sampling on the RGB luminance image; Converting the subpixel sampled (SPS) RGB luminance picture into an SPS-OCD luminance channel; Dividing the SPS-OCD luminance picture into separate SPS luminance and SPS chrominance channels; Performing high pass filtering on the SPS chrominance channel; Performing low pass filtering on the initial chrominance channel of the high resolution OCD picture; Performing subsampling on the filtered initial chrominance channel; Combining the subsampled lowpass filtered chrominance channel with the highpass filtered SPS chrominance channel; Combining the combined chrominance channel with the SPS luminance channel to form an error-reduced OCD low resolution image; And Converting the error reduced OCD low resolution image into an error reduced low resolution RGB image Method comprising a. A method of converting a high resolution image into a low resolution image with reduced visual error, Dividing a high resolution complementary domain (OCD) picture into separate initial luminance and initial chrominance channels; Generating sub-color domain (ACD) luminance images by performing subpixel sampling on the initial luminance channel; Converting the ACD luminance picture to an OCD luminance picture; Dividing the OCD luminance picture into individual subpixel sampled (SPS) luminance and SPS chrominance channels; Performing high pass filtering on the SPS chrominance channel; Performing lowpass filtering on the initial chrominance channel; Performing subsampling on the filtered initial chrominance channel; Combining the subsampled lowpass filtered chrominance channel with the highpass filtered SPS chrominance channel; And Combining the combined chrominance channel with the SPS luminance channel to form a low resolution image with reduced error Method comprising a. A system for converting a high resolution image to a low resolution with reduced visual error, Dividing means for dividing a high resolution complementary domain (OCD) image into separate initial luminance and initial chrominance channels; Subpixel sampling means for performing subpixel sampling on the initial luminance channel to generate an additive color domain (ACD) luminance image; A converter for converting the ACD luminance picture into an OCD luminance picture; Dividing means for dividing the OCD luminance picture into individual subpixel sampled (SPS) luminance and SPS chrominance channels; A high pass filter for performing high pass filtering on the SPS chrominance channel; A lowpass filter for performing lowpass filtering on the initial chrominance channel; Subsampling means for subsampling the filtered initial chrominance channel; A first combiner for coupling the subsampled lowpass filtered chrominance channel with the highpass filtered SPS chrominance channel; And A second combiner for combining the combined chrominance channel with the SPS luminance channel to form a low resolution image with reduced error System comprising a. A computer-readable recording medium comprising instructions for converting a high resolution image into a low resolution image with reduced error, wherein the instructions include: Dividing a high resolution complementary domain (OCD) picture into separate initial luminance and initial chrominance channels; Perform subpixel sampling on the initial luminance channel to generate a false color domain (ACD) luminance image; Convert the ACD luminance picture to an OCD luminance picture; Partition the OCD luminance picture into separate subpixel sampled (SPS) luminance and SPS chrominance channels; Perform high pass filtering on the SPS chrominance channel; Perform lowpass filtering on the initial chrominance channel; Perform subsampling on the filtered initial chrominance channel; Combine the subsampled lowpass filtered chrominance channel with the highpass filtered SPS chrominance channel; Combining the combined chrominance channel with the SPS luminance channel to form a low resolution image with reduced error And a computer readable recording medium. delete

Images