Object saliency detection is a significant and hot research issue in the field of image processing and computer vision, and has been widely used in many applications, such as objection recognition and detection, image search, image classification and machine vision, etc. Followed, adjusting the control parameters of algorithm is also a kind of important issues in the area of multimedia computing. This thesis focuses on the co-saliency object detection based on superpixel grid, and studies automatic adjustment method of the control parameters of target algorithm using nested differential evolution. The main contributions of this thesis are summarized as follows.1. This thesis proposes a fast and high accuracy approach to detect the co-saliency object region. The proposed approach uses the high-dimension integral matching method based on superpixel grid of image to detect the co-saliency object from the image pair. Among the approach, the method uses the punishment map to assess local distinctness, and by solving the global energy function to detect the common saliency object regions from image pair.2. This thesis proposes a method for automatic adjustment of the control parameters. Because of the existence of the parameter configuration problems in the saliency detection methods, that is, different setting algorithm parameters affect the merits of the algorithm results, so this thesis studies the automatic adjustment of the control parameters based on nested differential evolution. The purpose of this method is to adjust the control parameters of the target algorithm to automatically generate a set of optimal control parameters to the target algorithm.The effectiveness and practicality of the proposed approach are intensively evaluated with computer simulation and practical application experiments. |