Dissertation > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device

Research on Image Segmentation Based on Visual Saliency Model and Graph Cuts

Author LiuHongCai
Tutor HuYingSong
School Huazhong University of Science and Technology
Course Applied Computer Technology
Keywords Visual Saliency Model Graph Cuts Image Segmentation
CLC TP391.41
Type Master's thesis
Year 2013
Downloads 11
Quotes 0
Download Dissertation

Image segmentation is a basic computer vision processing technology, which is the foundation of pattern recognition and image analysis. Effective and reasonable image segmentation can extract valid information as much as possible, so that a higher level of image understanding becomes possible. With the development and application of computer, image segmentation has been widely used in the medical, transportation, industrial automation, video image processing. All in all, the image segmentation has important research significance, image segmentation, good or bad, will directly affect the subsequent image processing.First, describes several commonly used current image segmentation algorithm, and compare their advantages and disadvantages. Focuses on the visual saliency model and graph-based image segmentation method. Visual saliency model can quickly find out the significant regions in the image, but can not accurately cut the salient region out completely from the image. Graph-based image segmentation method, which is based on energy minimization framework, has quick segmentation speed and good segmentation results, but only in the case of human interaction, it can realize its advantages. Combining with the advantage of the visual saliency model, presented an image segmentation method based on visual saliency model and graph cuts. It uses visual saliency model to get the saliency regions in the image, which are used as initial conditions in the later segmentation. It not only maintains the fast speed of graph-based segmentation, and also achieves an automatic segmentation by using visual saliency model.Through a large number of experiments, obtained more satisfactory parameters combinations, which may be affecting the performance of the method. At the same time, compared to the common graph-based algorithm, obtained a fully automatic segmentation, as well as had a good segmentation performance and fast segmentation speed in the condition of a similar segmentation results.

Related Dissertations
More Dissertations