The Saliency Detection And Application Based On Color Contrast And Statistical Characteristics Of The Image Cell

In the process of visual saliency detection, as there’re different ways to measure theoptical parallax, the results of different saliency detection algorithms have hugedifferences. How to meet the demands of practical application according to the principleof vision salient image is worth further study. In this paper, we study a new algorithm forthe visual saliency detection, which uses the image region contrast and spatial distributionfeatures, that makes the shape and edge of the salient image much more clear and morefits the requirements of practical application according to the content of the image.Firstly, from the biological visual attention mechanism, we analyze the rationalityand feasibility of the image-cells-based saliency detection algorithm; according to theprobability distribution of the image’s color feature vector and spatial feature vector, wechoose a new algorithm to construct the image cell which is improved and self-adaptiveMean-shift image segmentation.Secondly, we summarize the common visual characteristics of saliency detectionalgorithm and analyze the principles of the image contrast feature, using the image cellcontrast to make saliency detection. While we cluster the colors of image cells, calculatethe image’s main color analyses the spatial distribution characteristic of the image’s maincolor, according to the variance of the clustering colors’ spatial distribution to detect thesaliency.Thirdly, to calculate the key salient region, we combine the test results of two modelthrough the Gaussian model to reach the final salient region detection result. Comparedwith existing methods, the algorithm considers the salient influence factors on color andspace, the generated result has better precision and recall rate, it can obviously inhibit thecomplex texture and noise, remove the influence of complex background and verify thefeasibility of the algorithm on the public test data sets.Finally, this paper summarizes and expands the application of the visual saliencybased on the visual saliency detection. Using the saliency detection algorithm to solve theproblem of image intelligent scale and image stylized, we further verify the effectiveness of the proposed algorithm.