# US 2014/0153843 A1 · Jagged edge reduction using kernel regression > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 2014/0153843 A1 - Document ID: US20140153843A1 - Status: Published application - Kind: A1 - Application number: 13/690,783 - Filed: 2012-11-30 - Published: 2014-06-05 - Earliest claimed: 2012-11-30 - Assignees: Sharp Laboratories of America, Inc. - Inventors: Mohammad Aghagolzadeh; Christopher A. Segall - Web page: https://patentgrep.com/patent/US20140153843A1 ## CPC - G06T 5/73: Deblurring; Sharpening - G06T 3/40: Scaling of whole images or parts thereof, e.g. expanding or contracting - G06T 3/403: Edge-driven scaling; Edge-based scaling - G06T 5/20: using local operators - G06T 7/13: Edge detection - G06T 2207/10004: Still image; Photographic image - G06T 2207/10024: Color image - G06T 2207/20016: Hierarchical, coarse-to-fine, multiscale or multiresolution image processing; Pyramid transform - G06T 2207/20024: Filtering details - G06T 2207/20192: Edge enhancement; Edge preservation - G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL ## Abstract An image decoder includes a base layer to decode at least a portion of an encoded video stream into a first image having a first image format. The image decoder can generate a color space prediction by scaling a color space of the first image from the first image format into a color space corresponding to a second image format. The image decoder includes an enhancement layer to decode the encoded video stream to generate a second image in the second image format based, at least in part, on the color space prediction. ## Claims ### Claim 1 A method comprising: determining, by an image processing system, directional gradients corresponding to pixels in an image; generating, by the image processing system, a linear predictor and a quadratic predictor from the directional gradients; and modifying, by the image processing system, the pixels in the image with kernel regression utilizing the linear predictor and the quadratic predictor. ### Claim 2 (depends on 1) The method of claim 1, wherein the directional gradients correspond to magnitudes of the linear predictor in an x-direction and a y-direction relative to the corresponding pixels in the image. ### Claim 3 (depends on 1) The method of claim 1, wherein the kernel regression utilizing the linear predictor and the quadratic predictor is configured to reduce curvature of isophotes in the image and maintain second-order curvature across the isophotes in the image. ### Claim 4 (depends on 1) The method of claim 1, further comprising: generating, by the image processing system, coherence values based on pixel variances in the image; and modifying, by the image processing system, the pixels in the image with kernel regression utilizing the coherence values to adjust a size of a kernel. ### Claim 5 (depends on 4) The method of claim 4, wherein the coherence values are configured to identify a measure of uniformity in luminance variations local to the pixels. ### Claim 6 (depends on 4) The method of claim 4, wherein the pixel variances correspond to magnitudes of luminance variations local to the pixels in directions of maximum luminance variations and directions orthogonal to the directions of maximum luminance variations. ### Claim 7 (depends on 6) The method of claim 6, further comprising: comparing the magnitudes of luminance variations local to the pixels with threshold values; and setting the magnitudes of luminance variations local to the pixels to a preset value when the magnitudes of luminance variations local to the pixels exceed the threshold values. ### Claim 8 (depends on 4) The method of claim 4, further comprising: determining to skip modification of selected pixels based on the corresponding coherence values in relation to the corresponding pixel variances in the image. ### Claim 9 An apparatus comprising: an image processing system configured to generate directional gradients corresponding to pixels in an image, generate a linear predictor and a quadratic predictor from the directional gradients, and modify the pixels in the image with kernel regression utilizing the linear predictor and the quadratic predictor. ### Claim 10 (depends on 9) The apparatus of claim 9, wherein the directional gradients correspond to magnitudes of the linear predictor and the quadratic predictor in an x-direction and a y-direction relative to the corresponding pixels in the image. ### Claim 11 (depends on 9) The apparatus of claim 9, wherein the kernel regression utilizing the linear predictor and the quadratic predictor is configured to reduce curvature of isophotes in the image and maintain second-order curvature across the isophotes in the image. ### Claim 12 (depends on 9) The apparatus of claim 9, wherein the image processing system is further configured to generate coherence values based on pixel variances in the image, and modify the pixels in the image with kernel regression utilizing the coherence values to adjust size of kernel. ### Claim 13 (depends on 12) The apparatus of claim 12, wherein the coherence values are configured to identify a measure of uniformity in luminance variations local to the pixels. ### Claim 14 (depends on 13) The apparatus of claim 13, wherein the pixel variances correspond to magnitudes of luminance variations local to the pixels in directions of maximum luminance variations and directions orthogonal to the directions of maximum luminance variations. ### Claim 15 (depends on 12) The apparatus of claim 12, herein the image processing system is further configured to compare the magnitudes of luminance variations local to the pixels with threshold values, and set the magnitudes of luminance variations local to the pixels to a preset value when the magnitudes of luminance variations local to the pixels exceed the threshold values. ### Claim 16 A method comprising: detecting edges in an image; and selectively applying kernel regression to portions of the image based, at least in part, on the detection of the edges in the image. ### Claim 17 (depends on 16) The method of claim 16, wherein selectively applying kernel regression to portions of the image further comprises: dilating the detected edges of the image to include portions of the image adjacent to the detected edges in a dilated image region; and selectively applying kernel regression to the portions of the image within the dilated image region. ### Claim 18 (depends on 16) The method of claim 16, wherein detecting edges in the image further comprises comparing the image to jagged edge configurations, and wherein selectively applying kernel regression further comprises applying kernel regression to the portions of the image corresponding the jagged edge configurations. ### Claim 19 (depends on 18) The method of claim 18, wherein the comparing of the image to jagged edge configurations is performed by a matched filter bank loaded with the jagged edge configurations. ### Claim 20 (depends on 18) The method of claim 18, wherein selectively applying kernel regression further comprises adjusting a size of kernels utilized in the kernel regression based on the comparison of the image to jagged edge configurations.