Illumination correction method for image

Abstract

The invention provides an illumination correction method for an image. The illumination correction method comprises the steps of step 1) estimating an illumination component of the image by using median filtering; step 2) determining whether the illumination component of the image is uniform, executing step 3) on the non-uniform illumination component, and ending illumination correction on the image for the uniform illumination component; step 3) detecting whether an interference region exists in the illumination component of the image, and if yes, removing the interference region from the illumination component of the image; and step 4) calculating a reflection component of the image according to the image and the illumination component of the image, and taking the reflection component as an illumination-corrected image. According to the method, the illumination component can be accurately estimated, meanwhile, details of the image can be well reserved, the pressure of post-processing of the image is reduced, and the method is suitable for illumination correction of the image with the high fineness requirement.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to an illumination correction method for images. Background technique [0002] In real life, the quality of images taken and collected under various lighting conditions is uneven, and it will affect subsequent ORC recognition, defect detection, etc. To solve this problem, researchers have proposed methods such as Retinex image enhancement method, grayscale transformation method represented by histogram equalization method, homomorphic filtering method based on illumination-reflection model, and gradient enhancement method to correct the illumination of images. Among them, the Retinex image enhancement method has become the mainstream illumination correction method due to its advantages of good processing effect and fast speed. [0003] The basic principle of the Retinex image enhancement method is that the original image is the product of the illumination component...

AI Technical Summary

Problems solved by technology

However, this multi-scale Retinex image enhancement method also has the following problems: because the Gaussian surround function is used to estimate the illumination component, and the Gaussian surround function has a physical meaning similar to the mean filter, its goal is to smooth the image and compared with the nonlinear filter The smoothing effect is more moderate, so it is easy to cause the loss of image detail information, such as the loss of texture information, insufficient edge sharpening, etc.

When using the multi-scale Retinex image enhancement method for illumination correction, it may affect the later image processing (such as defect detection, ORC recognition, etc.) due to the loss of image details

Images