Image white balance method based on sparse representation

Abstract

The invention discloses an image white balance method based on sparse representation. The method comprises the following steps of obtaining a large number of training images under various scenes and environments, utilizing a standard color pallet for conducting measuring to obtain an illumination chromaticity value of each training image in the process of shooting, conducting statistics on the corresponding color feature of each training image, giving a test image, utilizing the illumination chromaticity values and the color features of the training images to estimate the illumination chromaticity value of the test image, utilizing the obtained illumination chromaticity value of the test image, correcting the color of each pixel in the test image through a diagonal model, and obtaining the image after the white balance. The method can be applied to the fields of white balance processing of a camera, illumination transformation of the image, color constancy calculation and the like, and has the wide application prospect.

Description

technical field [0001] The invention relates to the technical fields of image processing, computational photography, pattern recognition and the like, in particular to an image white balance method based on sparse representation. Background technique [0002] As a simple, direct and effective feature, color has been widely used in various computer vision-related fields such as image processing, object recognition, image retrieval, and scene understanding. However, color is an extremely unstable image feature. The image color obtained by any imaging device depends on at least three main factors: the physical reflection characteristics of the object surface in the scene, the lighting conditions in the scene during imaging, and the imaging conditions. The imaging parameters of the device, therefore, there may be huge differences in the image color of the same scene under different lighting. Image white balance technology is to eliminate the influence of light on color, and cor...

AI Technical Summary

Problems solved by technology

However, the biggest disadvantage of these two algorithms is that the estimation model is fixed, that is, given the training data set, after using the prediction model obtained by these two methods, all test samples are illuminated using this fixed model.

This fixed model often constrains the generalization ability and adaptability of the algorithm

Images