基于纹理分离和美学评价的图像显著性区域检测技术研究-软件工程专业论文.docxVIP

优秀毕业论文 精品参考文献资料 AbstractAbstract Abstract Abstract With the rapid development of Intemet technology and multimedia technology,the accumulation of images data grow exponentially．How to efficiently and intelligently deal谢th these image data has become a hot research．In recent years， the image saliency region detection becomes more and more mature as a direction of computer vision．Image saliency region detection means that computers automaticlly mark salient regions that cause human visual attention by simulating the human visual characteristics．Thetechnology allows automatic identification of objects in natural images，generating grayscale image or a binary image for subsequent processing operations，such as image scaling for context-sensitive，image retrieval，image obj ect recognition and image cropping．Thus，the study of salient region detection has a very important significance． We did a lot experimental study of the existing image saliency region detection technology and found that the core object detection would be disturbed when the background has a complex and prominent texure．And the current algorithms are not suitable for catoon figure detection．To solve these problems，the work this paper has done is as following： First，to deal with the natrual pictures with cluttered background，we propose a texture separation algorithm for saliency detection．The algorithm separate texels while maintaining the main structure and reduce the prominence of the texels． Algorithm reduce the effect of background on the subject and achieve beRer result． Second，the paper takes the more and more cartoon images as the research object， analyze the basic information and propose a salient region detection algorithm based on aesthetic evaluation principles．It is a way to evaluate the aesthetic value of images based on some composition principles．As we all know,the important figures are placed at the prominent position by professionals．The algorithm combine the aesthetic information，gradient