A single image rain removal method

Abstract

The invention discloses a method for removing rain from a single image. First, a rainy image is decomposed into a high-frequency image and a low-frequency image based on bilateral filtering. Then, in the dictionary learning stage, in order to reduce the similarity between rainy atoms and non-rainy atoms, the incoherence of the dictionary is introduced, and a new objective function is constructed, which can not only ensure the reliability of the rainy dictionary and the non-rainy dictionary when clustering , and the learned incoherent dictionary has properties similar to tight frames, which can approximate isometric tight frames. Through the sparse representation of the high-frequency image with the rain dictionary and the rain-free dictionary, the rain component and the non-rain component in the high-frequency image can be better separated, and the high-frequency non-rain component and the low-frequency image are superimposed to realize image deraining. Experimental results show that the non-coherent dictionary learned by the method of the present invention has better sparse representation performance, less rain lines remain in the image after rain removal, edge details are better preserved, and the visual effect is clearer and more natural.

Description

technical field [0001] The invention relates to the field of image processing, in particular to a method for removing rain from a single image. Background technique [0002] In rainy weather, raindrop particles are generally larger than 100 μm and are easily captured by the lens. Therefore, due to the influence of rainy weather conditions, some texture and detail information of images captured by outdoor lenses are easily blocked by rain lines, resulting in problems such as excessively bright local areas and blurred background images. The degradation of image quality in rainy days greatly restricts the functions of outdoor intelligent vision systems such as visual surveillance, visual navigation and target tracking. Moreover, the state of raindrop particles is changeable, and the direction and thickness of rainlines are different in different situations. Therefore, it is of high research and application value to study how to recover high-quality images from various rainy de...

AI Technical Summary

Problems solved by technology

Huang et al. used the affine propagation method to realize the clustering of dictionary atoms. Each type of dictionary atom contains similar edge and texture information, and the high-frequency no-rain component is obtained by using the large variance of the rain-free atom class, and then superimposed with the low-frequency image to realize the image To remove the rain, this method improves image edge preservation, but there are certain rain lines remaining, and the dictionary clustering method leads to poor real-time performance of the algorithm

Luo et al. proposed a nonlinear rain image model using the color filtering principle in PS (Phtoshop), and used discriminative sparse coding to realize image deraining. This method constrains the relationship between the rain-free image and the sparse representation coefficient of the rain component. Realize the image to remove the rain, the phenomenon of residual rain line is serious, and the rain is not completely removed

[0005] Image deraining based on dictionary learning and sparse representation has achieved certain results, but there are still the following problems: (1) There is overlap between the rain line and the background image texture in the rainy image, and the dictionary obtained by the existing dictionary learning method The similarity between atoms is high

The orthogonal base coherence is equal to 0, but the orthogonal base cannot guarantee the sparsity of the sample representation

Images