Content-Based Image Retrieval Using the Local Structures of Color and Edge Orientation

In this paper, we propose a simple, yet very powerful image representation, namely local structure pattern descriptor, to describe the local structure features of edge orientation and color for image retrieval. First, it converts color image from RGB color space to Lab color space, and then colors information and edge orientation are extracted in Lab color space, respectively, and then five textons types be used to detecting the local structures of edge orientation and color. Finally, a new algorithm is proposed to represent image features. The proposed algorithm is extensively tested on two Corel datasets with 15,000 natural images. Image retrieval experimental results have shown that the performance is better than that of the local binary pattern histogram and Gabor filter method significantly. The local structure pattern descriptor has the good discrimination power of color, edge features and certain spatial features.