Visual Significant Computing And Applied Research In Natural Images Of The Object Of Interest Is Detected

Visual attention means that when human beings face the complex scenes, they will focus the important area in the scene quickly, and handle it in priority. Nowadays, Image processing takes a lot of time and computing power when dealing with the complex scenes. Therefore, it is very necessary to imitate the Human Visual System to establish a visual model, which has important applications in Interested Object Detection, Object Recognition, Image Retrieval, Image and Video Compression etc. So it is very necessary to research it.In the field of computer vision, the research of visual attention is becoming a hot spot. So we propose two visual computing methods on the basis of existing models, and use it in the interested object detection for natural images. The main contents of this paper are as follows:(1) The status of visual saliency is stated in detail, the advantages and disadvantages of existing models are analysed, and the basics of visual saliency is explained.(2) Propose a visual saliency method based global contrast in connection with the salient feature maybe different through the different color component. According to the color of salient regions is rare in the global picture, and it is centralized in the spatial distribution. We obtain the salient picture both in the HSV space and the Lab space. And then merge them to obtain the final salient image. Then we use a large number of images to analyse the method, the result indicates that this method is better than several typical methods.(3) Propose a visual saliency method based the candidate area in connection with the disadvantage of obtaining the salient area, which depend on the salient picture totally. We use the method of hierarchical image segmentation to generate candidate region, and computing the salient value through the spatial domain, frequency domain, closed contours and the local color contrast. Then we get the salient area through sorting the candidate area. We use a large number of pictures to analyse the method. And the result indicates that this method is better than several typical methods.(4) Through improve the visual saliency models, we use them in the field of object detection. Firstly we divide the natural image into six classes:Sky, River, Desert, Grass, Forest and City background. We make an experiment and analysis the method. The results show the effectiveness of our method.