关键词: 显著性检测, 模糊现象, 多标签传播, 边缘局部分析

Abstract: Aiming at fuzzy phenomena around edge of salient object in the saliency detection, a new image saliency detection algorithm based on multi-label propagation and local analysis of edge is proposed. Firstly, the most reliable object and background labels are extracted, then, the salient objects have been located accurately with the multi-label propagation algorithm based on cells updating strategies. A novel strategy is proposed for further optimization, ?which analyzes local clues of the extracted edge labels and accomplishes nodes classification to fall within the selected local range. Therefore, the fuzzy phenomena are eliminated around salient object edge. The experimental results on the public datasets show that the proposed algorithm can suppress the backgrounds around the salient object effectively, solves fuzzy phenomena problem around the salient object completely and highlights the salient objects clearly and uniformly.

Key words: saliency detection, fuzzy phenomena, multi-label propagation, local analysis of edge