Image rain removal method based on learning convolutional sparse coding

Abstract

The invention relates to the technical field of image restoration, and discloses an image rain removal method based on learning-type convolution sparse coding. S1: obtains an image to be processed; S2: uses a comprehensive-based global and local gradient prior to characterize the background image, and Use learning convolution sparse coding to process rain marks, and construct a comprehensive global, local gradient and learning convolution sparse coding image rain removal model; S3: Solve the comprehensive global, local gradient and learning volume in step S2 Integrate the sparsely coded image to remove the rain model, and output the result to remove the rain image u and rain layers infographic r . Compared with the prior art, the comprehensive global and local anisotropy prior knowledge of the present invention protects the background target information, and the learnable convolution sparse expression represents the variable-scale rain mark target, so that the rain mark can be better carried out. Detection, not only can effectively detect the rain mark information to achieve the target to remove the rain, but also make the background image information after the rain removal also be better protected.

Description

technical field [0001] The invention relates to the technical field of image restoration, in particular to an image rain removal method based on learning convolutional sparse coding. Background technique [0002] At present, photoelectric imaging instruments are widely used in the field of information perception. However, the current imaging system is mainly designed in a relatively ideal environment, that is, the working environment of the imaging system needs to have good lighting conditions and good weather conditions. When the imaging system is affected by bad weather such as rain and snow, its information perception performance is greatly affected, and the obtained image is affected by the traces (ie, rain marks) produced by the rain falling process. Although experts and scholars have proposed a variety of methods for removing rain marks from images, such as methods based on deep learning neural network, generative adversarial neural network, rain target structure prio...

AI Technical Summary

Problems solved by technology

However, since rain is an uncontrollable natural factor, there are still some rain marks in the image after these methods remove the rain marks.

Images