The Extraction and Application of the Color Texture Feature Based on Quaternion Gabor

Abstract

The Gabor is a standard means to extract texture feature of image, besides it was used to image classification and image segmentation far and wide. But, traditional methods that extracting texture feature always ignoring the color information and losing dependency between one channel and others. That makes it difficult to remain the raw information of image. To address above problems, this paper proposed a quaternion Gabor method to extract color texture feature. Firstly, according to the traditional Gabor filter and quaternion Euler’s formula, the quaternion Gabor filter was determined. Then, the multi-scale, multi-direction color texture image was obtained by quaternion Gabor convolution algorithm. Finally, the Tamura feature was extracted from the feature image to test the proposed method. Experimental results show that the proposed method can retain coarseness, contrast and directionality to a great extent, the feature image if better than traditional Gabor feature image and LBP feature image in retaining the Tamura texture feature. Besides, the quaternion Gabor can also use in image enhancement. The second experiment shows the application on image enhancement.