An Orientation Independent Texture Descriptor for Image Retrieval

In this paper, a new orientation independent texture descriptor based on the relation characteristic of pixels in their local neighbor is proposed, which is termed quantized compound change histogram. For every pixel in the image, the compound rate of change of gray value of the pixel in four different directions in its neighbor is computed and is quantized into 40 bins to get the quantized compound change at the pixel. Calculating the distribution of the quantized compound change over the whole image, the quantized compound change histogram is obtained. It is invariant to image rotation and translation. Experimental results show that the texture descriptor has better descriptive power for natural color images.