A new combination feature of S-LBP and main color descriptor for video frame retrieval

Abstract

In this paper, we draw Local Binary Pattern (LBP) and main color descriptor into content-based video indexing and retrieval. LBP is a powerful texture descriptor for its tolerance against illumination changes and its computational simplicity. Main color descriptor is used for representing main color distribution of the image. We propose a simplified LBP (S-LBP) operator as the texture descriptor. In most cases, texture distributions that ignore the actual positions or spatial arrangement in the image are not sufficient for representing an image. Therefore, we use zigzag based Run-length coding principle to produce S-LBP patterns histogram. Then we apply the combination of S-LBP feature and main color descriptor to represent an image in video retrieval. The experiment that to retrieval which video a given frame comes from demonstrates the combination method is efficient and effective.