Saliency Detection Algorithm under Complex Background
Xin Pang · Journal of Information and Computational Science · 2015
Recent years, saliency detection algorithm which is based on the characteristics of human visual attention gets more and more attention of researchers. A rapid, accurate, suitable for all kinds of background and robust detection algorithm has always been the goals pursued by majority of researchers. The significance of this article is to improve a good detection algorithm, and to make up for it under the complex background which cannot detect the significant goal well. This method first detects saliency area, extracts it from the input image, removes the most complex background, and then, on the extracted region detects the significant target again. For an extreme case, part of the significant target in the subimage boundary is extracted from the whole image, leading to a result that when detecting the significant goal for a second time, the parts on the subimage boundary will be considered as background which seriously affects the results. To solve the problem we magnify the background appropriately. The experimental results show that by improving the algorithm, it can extract the target accurately under relatively complicated background.