A study on the computational color constancy algorithm based on ANFIS-LSSVM
Computing color constancy refers to the ability to eliminate the influence of scene light sources and thus reproduce the true color of an object.Currently,significant improvements in color constancy accuracy have been achieved by using deep neural networks,but most deep learning algorithms have long training time,high computational complexity,and require a large number of training samples.To address this problem,this paper proposes a simple and effective method that combines adaptive neuro-fuzzy inference system(ANFIS)and least squares support vector machine(LSSVM).The method is divided into two phases:training and prediction.In the training phase,image features are first extracted to train two initial light source estimation models,ANFIS and LSSVM,respectively.Then the kernel function transform is used to fuse the two models,and the multivariate linear regression model of light source estimation is obtained by using the reserved training samples for further training.In the prediction phase,after extracting the features of the test image,the final color values of scene light source of the test image is predicted directly from the training data.The experimental results show that,compared with the deep learning method,the proposed method is of lower computational complexity and even with small training samples,and it also has good light source estimation performance.
computational color constancyilluminant estimationadaptive-network-based fuzzy inference system(ANFIS)least squares support vector machine(LSSVM)
NETL
NSTL
万方数据
维普
