Adjacent Local Binary Patterns Based on Color Space Fusion for Color Image Classification

Abstract

In this paper, we propose an improved image feature descriptor based on Local Binary Pattern, which is\r called Adjacent Local Binary Patterns based on Color Space Fusion (ALBPCSF). The proposed method\r fuses color feature and spatial relations. ALBPCSF uses the channel values of RGB and HSV color spaces\r to calculate the color feature. Then the proposed method considers the spatial relations which will be\r combined with the color feature. Finally, an image classification system framework based on ALBPCSF\r is given. In order to validate the performance, our method is compared with previous methods on Corel\r 1000 and MIT Vision Texture datasets. The results show that our approach is superior than other\r methods in color image classification.