Watershed-Based Texture Image Retrieval

The content-based image retrieval (CBIR) is a hot topic recently. In this paper, a novel algorithm, namely a watershed-based texture image retrieval algorithm, is proposed. The algorithm mainly consists of three parts. Firstly, after reduced the noise by the open-closing by reconstruction, the image is segmented into regions by an improved watershed transformation. Secondly, the segmentation regions are re-arrayed from big to small under pixel number, and selected from number one to number T-1. The remaining regions are combined to generate the region of order T. After above optimizing, the textural features regions are extracted to compose a feature vector of image based on color co-occurrence matrix. Finally, the similarity of two images will be determined by the similarity between texture feature vectors. Experiment results show that the proposed algorithm is efficient.