关键词: 多分辨率, 最小障碍算法, 梯度融合

Abstract: The saliency region of natural image is generally located in the center. The saliency region can be obtained by calculating the distance between the region and the boundary. Based on the above phenomena, many algorithms with high accuracy or fast running speed are generated. However, for improving the efficiency and the performance of the algorithm in order to realize real time detection, it needs to do further research on obtainning faster and more accurate saliency detection. To handle the above problems, a saliency detection algorithm based on multi-resolution minimum barrier and gradient fusion is proposed. Firstly, the multi-resolution image is formed by multiple sampling, and the saliency detection algorithm based on the improved minimum barrier is introduced to caculate the saliency of the processed image. Secondly, during the operation of the algorithm, gradient analysis is performed on the results of the algorithm and the background image. The two images are fused to improve the problem of the fuzzy region. After experimenting on multiple datasets, the algorithm can achieve the detection rate of 75 f/s while the correctness is above 90%.

Key words: multi-resolution, minimum barrier algorithm, gradient fusion