Image processing method and device based on color restoration, pedestrian recognition method and device and readable storage medium

Abstract

The invention provides an image processing method and device based on color restoration, a pedestrian recognition method and device and a readable storage medium. The image processing method comprisesthe following steps: 1) extracting continuous frames; 2) performing Gaussian blur; 3) performing channelization treatment; 4) performing quantization processing to obtain a reflection image; 5) performing weighted calculation to obtain an enhanced image; and 6) performing mean value processing and normalization on the enhanced image to obtain a preprocessed image, and outputting the preprocessedimage. According to the image processing method and device based on color restoration, the pedestrian recognition method and device based on color restoration and the readable storage medium, the enhanced reflection image can be more scientifically calculated, a recognizable image is clearly restored from a low-brightness image in a dark scene, and noise points are fewer than those of other imageenhancement algorithms. Meanwhile, different from a traditional linear and non-linear method which can only enhance a certain kind of features of the image, the method of the invention can achieve balance in three aspects of dynamic range compression, edge enhancement and color constancy, and therefore adaptive enhancement can be carried out on various images with different brightness.

Description

Technical field [0001] The present invention relates to the technical field of image recognition, in particular to an image processing based on color restoration, a pedestrian recognition method, device and readable storage medium. Background technique [0002] In the current security field, the technology to realize face and pedestrian recognition through image algorithms has gradually matured. However, the recognition of dark scenes with insufficient light has always been a problem of serious decline in accuracy. In order to solve the problem of accuracy loss caused by insufficient brightness, people have tried some traditional image enhancement algorithms such as histogram equalization, Gamma correction, and Laplacian. Since traditional image enhancement algorithms generally increase or decrease the brightness and darkness of the picture as a whole, when they are applied to more complex scenes such as surveillance images under street lights at night, there are both very brigh...

AI Technical Summary

Problems solved by technology

This makes the Laplacian operator as non-adaptive as Gamma correction, and the sampling of the operator template will directly affect the effect of image enhancement

[0010] Therefore, in actual security application scenarios, especially for pedestrian recognition in dark environments, the enhancement effect of existing algorithms is not even as good as that of the original image without image enhancement, such as continuing to perform detection and recognition tasks after image enhancement

Images