Natural image shadow elimination method

Abstract

The invention discloses a natural image shadow elimination method. The method includes the steps of extracting a shadow area through an interaction sample and the image matting thought, decomposing an input image into a shadow block set and a non-shadow block set through the shadow illumination distribution, finding a block with the texture most similar to the texture of each shadow block in the shadow area in the non-shadow block set to serve as the matched block of the shadow block, eliminating shadows in the shadow blocks through illumination recovery operators in the method according to information of the matched blocks, conducting the weighted average operation on overlapped areas between the blocks so as to achieve the shadow elimination result that the areas between the blocks are smooth and the illumination of the blocks is consistent with the illumination of the surrounding environment, and conducting the restrained texture synthesis operation on the area where boundary treatment needs to be conducted so as to achieve a more natural shadow elimination result. The shadow elimination method is good in effect, and the illumination recovery operators in the method can be used for various applications.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to a method for eliminating shadows of natural images. Background technique [0002] As we all know, there will be shadows when there is light and occlusion. Although the shadow in the image can provide information for the shape of the object in the image, the position of the light source, the lighting situation, etc., the existence of the shadow will have an adverse effect on the research of image-related problems, for example, because the shadow covers the part of the target object in the image Information, which reduces the recognition accuracy of the object; shadows will increase the complexity of image matching, target segmentation and tracking in image post-processing. In addition, for aesthetic considerations, it may be desirable to eliminate some shadows in the image that are not related to the image content but will affect the image content itself in t...

AI Technical Summary

Problems solved by technology

At present, although the shadow removal method based on the illumination invariant feature image and solving the Poisson equation, and the shadow removal method based on the statistical model have achieved good results in processing simple shadow images, but for complex image shadows and inconsistent shadows, the shadow removal method Results still need to be strengthened

Images