Texture Image Classification with Improved Weber Local Descriptor

摘要

Texture features play an important role in image texture classification. Inspired by Weber’s law, Weber Local Descriptor (WLD) has been proposed for image texture classification. Orientation component in Weber Local Descriptor is the gradient of an image, which does not properly represent the local spatial information of an image. In this paper for orientation component, we propose to compute the histogram of gradient instead of the gradient of an image. The gradient of an image is computed, then image is divided in to small spatial regions named as cells and histogram of each cell is obtained. We have tested our proposed scheme on publically available texture datasets named as Brodatz and KTH-TIPS2-a, which shows that our proposed method can achieve significant improvement as compared to the state-of-the-art method like Local Binary Pattern, Local Phase Quantization and Weber Local Descriptor.