Classification of color textured images using linear prediction errors and support vector machines.
In this paper we present a novel method for pixel classification. The goal is to approximate the distribution of the two dimensional multichannel linear prediction errors in order to improve the performance of color texture image classification. This method computes the class membership probability for each pixel of a given image. This probability uses the concept of clique potentials in a finite neighborhood of the pixel. The support vector machine (SVM) classification is applied to predict the class of each pixel belonging to the foreground. And finally, we do further refinement by neighborhood-check to omit all falsely-classified pixels. The results of the method are also compared to those of a non parametric and parametric pixel classification method respectively KNN and Bayes classifier. For the proposed method and with different color spaces, experimental results show better performances in terms percentage classification error, in comparison with the use of a multivariate Gaussian law.
Color texture classification, Support vector machines, Color space comparison, Linear prediction error