Demosaicing method and demosaicing equipment

Abstract

An artificial neural network is provided that is configured to convert a mosaic image that can be represented by pixels having color pixel values corresponding to a plurality of colors into a plurality of chromatic aberration images corresponding to the plurality of colors. A plurality of output color difference images corresponding to the plurality of colors are generated based on an input mosaic image that can be represented by pixels having color pixel values corresponding to the plurality of colors using an artificial neural network. A plurality of demosaiced color images corresponding to the plurality of colors are generated by summing an input mosaic image and each of the plurality of output color difference images pixel by pixel.

Description

[0001] Cross References to Related Applications [0002] This application claims priority to Korean Patent Application No. 10-2020-0138500 filed on October 23, 2020 at the Korean Intellectual Property Office (KIPO), the disclosure of which is hereby incorporated by reference in its entirety. technical field [0003] Example embodiments relate generally to image processing. More particularly, example embodiments relate to demosaicing methods and demosaicing apparatuses. Background technique [0004] Each pixel in an image sensor detects one color, and the other colors for each pixel can be obtained by interpolating the pixel values ​​of neighboring pixels. This color interpolation is called demosaicing. [0005] In general, demosaicing methods that reconstruct missing color pixel values ​​result in block noise, such as zipper effects, random color points, and the like. Also, there is a tradeoff between noise filtering and edge blurring. When interpolating pixel values ​​n...

AI Technical Summary

Problems solved by technology

Such problems are caused by, for example, averaging colored pixel values ​​across an edge when the colored pixel values ​​on either side of the edge differ greatly, and this averaging greatly reduces sharpness differences resulting in blurring

[0006] Most methods of demosaicing color filter array (CFA) images or mosaic images by linear interpolation can lead to artifacts such as reduced image detail, false colors, and aliasing

Because color channels in the current stage are restored using the color channels restored by the previous stage during multi-stage processing, errors may accumulate

Images