Image restoration method based on frequency band self-adaptive restoration model

Abstract

The invention relates to an image restoration method based on a frequency band self-adaptive restoration model, which is used for solving the problem that a reasonable structure and a fine texture of a damaged image cannot be reconstructed at the same time in the prior art. On the basis of a generative adversarial network model, an encoder-decoder network structure is used as a baseline network of a generator. In order to extract deep information from a known area, features of a damaged image are decomposed into low-frequency sub-bands and high-frequency sub-bands through discrete wavelet transform (DWT), and then features of the multi-frequency sub-bands are extracted through a convolutional network; and finally, each horizontal sub-band image is reconstructed through wavelet inverse transformation, and missing region information is complemented to generate an image. The idea of separately processing the low-frequency information and the high-frequency information is beneficial to pointedly processing the structure and texture information of the image. The image structure generated by the method is clearer, the texture is finer, and the restored image is more real and has no obvious boundary.

Description

Technical field: [0001] The invention relates to the field of computer image processing, in particular to an image repair method based on a frequency band self-adaptive repair model. Background technique: [0002] Image inpainting is a fundamental task in multimedia applications and computer vision, where the goal is to generate alternative global semantic structures and local detail textures for missing regions, and ultimately produce visually realistic results. It has been widely used in multimedia fields such as image editing, restoration and synthesis. Traditional patch-based image inpainting methods search and copy the best matching patch from known regions to missing regions. This traditional image inpainting method works well on static textures, but has limited effect on textures with complex or non-repetitive structures such as faces, and is not suitable for capturing high-level semantic information. [0003] In recent years, learning-based methods have modeled ima...

AI Technical Summary

Problems solved by technology

[0006] In order to solve the problem that the existing technology cannot reconstruct the reasonable structure and fine texture of the damaged image at the same time, the present invention provides an image repair method based on the frequency band adaptive repair model, the image structure generated by the method is clearer, the texture is finer, The repaired image is more realistic without obvious borders

Images