Image shadow reflection elimination method and device, computer equipment and storage medium

Abstract

The invention relates to an image shadow reflection elimination method and device, a computer device and a storage medium. The method comprises the steps of obtaining a to-be-processed image; performing Mosaic operation on the to-be-processed image, and inputting the to-be-processed image into the target detection model for target detection to obtain a target detection result; cutting the to-be-processed image according to the target detection result to obtain a high-definition image; inputting the high-definition image into an anti-generation network model to eliminate shadows and reflected light so as to obtain a restored image; and feedback of the restored image to the terminal. According to the invention, after the target detection model is adopted to carry out target detection on theimage to be processed, cutting is carried out according to the detection result so as to retain the image only having the target, and then shadow elimination and reflection elimination are carried outsimultaneously through the elimination model obtained based on PAIRWISE-GAN network training, so that the problems of reflection and shadow can be solved simultaneously; and the relation between theontology and the shadow shape of the ontology can be better fitted, the efficiency is high, and missing detection is not likely to happen.

Description

technical field [0001] The present invention relates to an image processing method, and more specifically refers to an image shadow reflection elimination method, device, computer equipment and storage medium. Background technique [0002] Shadows and reflections appear in various image processing scenarios in daily production and life, and they are an important and inevitable task for computer image algorithms. A good preprocessing image algorithm that can remove shadows and reflections to the greatest extent can greatly improve the stability and accuracy of subsequent algorithms, especially for tasks such as image segmentation and object detection and tracking. Existing algorithms based on pattern recognition rely on additional artificial preset information, and the processing of shadows and reflections is carried out separately. [0003] The existing shadow removal algorithms are mainly divided into two categories: feature-based shadow removal algorithms and model-based ...

AI Technical Summary

Problems solved by technology

Existing pattern recognition-based algorithms rely on additional artificial preset information, and the processing of shadows and reflections is performed separately

[0003] The existing shadow removal algorithms are mainly divided into two categories: feature-based shadow removal algorithms and model-based shadow removal algorithms. Feature-based algorithms mainly use the differences between shadows and background images in some basic attributes, such as brightness and hue. , saturation, gradient, texture and other features. This type of algorithm is simple in principle, but it is very sensitive to illumination and is prone to missed detection; the model-based algorithm is a shadow removal algorithm based on scene modeling. Scene modeling needs to know factors such as environment, target and lighting, and the calculation is more complicated, and the efficiency is relatively low

[0004] For shadow and reflection removal, the existing algorithm only performs a single shadow removal or reflection removal repair operation for the image, but cannot achieve the fusion operation of the two, resulting in the need for multi-step operations, resulting in redundant and complicated processes. , the traditional shadow removal and reflection operation will be affected by different degrees and different angles of light, which will form shadows in multiple areas and different reflection phenomena, which will seriously affect the accuracy of recognition. The shadow removal and reflection operation needs to be carried out through shadow or reflection detection first. The shadow or reflection removal operation requires a two-step strategy to repair the image. In this method, the removal operation in the second step often depends on the detection result of the first step. If there is an error in the detection in the first step, this will cause the repaired image to fail. Achieve very good results; the input information of the traditional shadow removal and anti-reflective image is RGB color, so the operation of eliminating its noise will lead to the model more fitting and color information, thus ignoring the shape relationship information between the body and the shadow

Images