Texture image retrieval using local binary edge patterns

Texture is a fundamental property of surfaces, and as so, it plays an important role in the human visual system for analysis and recognition of images. A large number of techniques for retrieving and classifying image textures have been proposed during the last few decades. This paper describes a new texture retrieval method that uses the spatial distribution of edge points as the main discriminating feature. The proposed method consists of three main steps: First, the edge points in the image are identified; then the local distribution of the edge points is described using an LBP-like coding. The output of this step is a 2D array of LBP-like codes, called LBEP image. The final step consists of calculating two histograms from the resulting LBEP image. These histograms constitute the feature vectors that characterize the texture. The results of the experiments that have been conducted show that the proposed method significantly improves the traditional edge histogram method and outperforms several other state-of-the art methods in terms of retrieval accuracy.