A Study On Visual Significantness Based On Semantics

As one of the most common information carrier in our daily life, image is an objectivedescription of the real world. The information human get in daily life are about75%from thevisual. And visual presentation in front of people is image or video, the study of the imageprocessing is more and more deeply at present. In the field of image processing, visual saliencyextraction as one of the integral part can imitate human visual attention to extract the partinformation that we focus on. The part is the key area of the image. To extract the saliency areaaccurately not only can help to let us know the cognitive rules of the human visual, but also canlay a solid foundation for other fields of the image processing.A better image will be more useful than many words. It is also more persuasive. But whenpeople see an image, they will only notice just a part of the image. So, we need to extract thepart that we are interested in, that is to extract the extracted the useful part of the image. Thesepart of the image is called interested areas.In this paper, we carried on the through research about the detection and extraction of theimage salient region. And in order to obtain the better effect of the detection and extraction ofthe image salient region, we combine patchnet and the image optimization to improve theexisting algorithms about extraction and detection of the image salient region.First of all, we introduce the research significance and the research status at home andabroad about the extraction and detection of the image salient region. And then we detailseveral classic visual attention models at home and abroad. Firstly, we introduced the classicalItti visual attention model. The principle of this algorithm is to process the image with thefeature images of it. It calculated the color feature of the image, the brightness feature of theimage and the location feature of the image. Finally, it fused these three feature image to get thefeature image of the image. Then, this character introduced the SR algorithm, it extracted thesignificant area of the image by extracting the residual spectrum. Finally, it introduced the RCalgorithm and the main idea of this algorithm is to integrate the spatial relations with thecontrast calculation of the regional level.With the development of the science and technology, visual saliency extraction algorithmis as one the most important research directions in the field of the image processing. In terms ofthe existing algorithms, most saliency detection algorithms can get better effect only for theimages with single salient object. However, because of the complicated varieties of the colors,textures and background, the saliency detection effect is not very ideal for the images withsimilar salient regions. In this paper, with the help of the existing visual saliency detectionalgorithms, we put forward a new algorithm to solve the saliency detection about the images with similar salient regions. We put forward an improved saliency extract algorithm combinedwith the patchnet technology. This algorithm used patchnet to record the context semanticrelations of the image. According to the following steps to detect the saliency:1. To detect theoutline of the saliency area;2. using the sketch matching method to find the location of saliencyobjects in the image;3. using the context Map to further extract saliency objects to get thefinally result. The finally experimental result showed that the algorithm with multiple imagesof the objects of similar significant testing can have very good effect in this paper.To extract the key area of the image correctly not only can greatly simplify the subsequentprocessing step, also can avoid wasting unnecessary computer resources. So on the basis of theexisting saliency area extraction algorithms, the fourth part in this paper proposes an imagesaliency extraction based on composition optimization. the idea is put the application of therules of composition optimization on the image saliency extraction. First of all, we introducethe research significance and the research status at home and abroad about the extraction anddetection of the image salient region. And then introduced several common image optimizationrules. And then we detail how to apply the image optimization rules on the image saliencydetection. The experimental results show the effectiveness of the algorithm in this paper.At the end of this article, the work to the full text is summarized and the future work isprospected.