Quaternion Graph Wavelet Transform for Color Texture Classification
Wei Cheng, Yulong Qiao · 2022 7th International Conference on Signal and Image Processing (ICSIP) · 2022
The utilization of color and texture information is an important research direction in image processing and pattern recognition. Quaternion graph wavelet transform can combine the structure and relationship between samples to provide richer information at different scales. Therefore, this paper proposes a color texture classification method based on quaternion graph wavelet transform. Specifically, the quaternion graph wavelet transform is used to decompose the color texture. Then the local quaternion singular value decomposition is performed on the subband coefficients to extract local texture information and the Weibull distributions are used to model effective singular values. Finally, experiments on two datasets using the nearest neighbor classifier measured by KL divergence demonstrate that the proposed method is effective for color texture classification.