Local Feature-based Saliency Detection Algorithm And Its Application In Image Retrieval

Image saliency detection is very useful in many field of image processing. Saliency detection is a comprehensive research involving image analysis, feature extraction and visual attention recognition. Image saliency detection can be applied to various areas, for example, content-based image retrieval. In traditional CBIR, extracting features from an image generates huge amount of information, which limits the processing time. It ignores that different regions of the image have different attractiveness to human visual system. In this paper, saliency detection is adopted to find the region of interest in image for image retrieval. Under the demand of image retrieval, this article researches the extraction of Regions of Interest, in order to improve the results of image retrieval. In this paper, the main works are as follows:(1)Common saliency detection algorithm calculate the contrast, to find the points which has the biggest difference with the surrounding. This process is similar to local feature points distribution which generated on feature detection stage. We introduce a salient region algorithm based on extraction of local feature. Utilize the SURF to generate the distribution matrix of feature points. The distribution matrix describes the distribution of feature points.(2)The greatest concentration of feature point is salient region. Utilize the idea of dynamic programming to compute the sum of max sub-matrix efficiently, which can detect saliency region of image efficiently.(3)Nonlinear gaussian distance function for similarity measure is introduced in CBIR based saliency detection.