Robust color texture retrieval method using co-occurrence matrix of pattern spectrums
Shota Takei, Shigeo Wada · 2012
In this paper, we propose a robust color texture image retrieval method using co-occurrence matrix of pattern spectrums (PSs). The proposed method has robustness to geometric distortions such as shift, rotation, scaling, skew, projection and their combinations. The feature of texture pattern and color is extracted as co-occurrence matrix of PSs of multiple component planes that can be robust to geometric distortion. To realize efficient calculation, binary pattern of RGB components are combined with grey level luminance component to obtain feature vector. In simulations, performance of the proposed method is analyzed under various distortion conditions with similar image retrieval system. Effectiveness is verified by evaluating retrieval accuracy rates. Retrieval accuracy is examined with multi-level as well as binary planes.