Image denoising method and device

Abstract

The embodiment of the invention discloses an image denoising method and device. The method comprises the steps of determining a denoising adjacent region and a reference adjacent region from a Bayer template image; on the basis of color components of a center pixel point, obtaining a first RGB color component value in the denoising adjacent region and a second RGB color component value in the reference adjacent region; on the basis of the color components of the center pixel point, conducting zooming processing on the first RGB color component value to obtain a first set of RGB color vectors, and conducting zooming processing on the second RGB color component value to obtain a second set of RGB color vectors; on the basis of the first set of RGB color vectors and the second set of RGB color vectors, obtaining a smoothing weight of the reference adjacent region relative to the denoising adjacent region; on the basis of the smoothing weight, conducting denoising processing on the denoising adjacent region.

Description

Technical field [0001] The invention relates to the field of image processing, in particular to an image denoising method and device. Background technique [0002] With the popularity of various digital instruments and digital products, images and videos have become the most commonly used information carriers in human activities. They contain a large amount of information about objects and become the main way for people to obtain original information from the outside world. However, in the acquisition of images, During the transmission and storage process, various noises are often interfered and affected to degrade the image quality. In order to improve the quality of the image, the prior art can use spatial domain neighborhood filtering methods such as nonlocal-average (nonlocal- mean) The algorithm performs denoising processing on the image, so that the noise of the denoising processed image is reduced, and the display effect of the processed image is improved. [0003] However,...

AI Technical Summary

Problems solved by technology

[0005] If v(N i ) is a rectangular neighborhood formed by 3*3 pixels centered on pixel i, v(N j ) is a rectangular neighborhood formed by 3*3 pixels centered on pixel j, v(N i ) and v(N j The square of the Euclidean distance of ) is the sum of the squares of the point differences of nine corresponding pixels in two blocks, but each pixel in each neighborhood rectangular block has only one color component, which cannot be well represented by 9 pixels of a single color component The distribution of details in the neighboring rectangular blocks makes it inaccurate to measure the similarity directly based on the square of the Euclidean distance between two neighboring rectangular blocks, which will lead to inaccurate selection of similar pixels to participate in the weighted filter. The problem of poor noise effect

Images