Research On Scale Space Image Analysis And Its Applications

Scale space image analysis and its applications are the hotspot and difficulty in image understanding area. Adaptive image display is a kind of visual information expression. The process fits the images for ours watching, via adjustment environmental parameters. The main part of automatic target detection is visual selective attention. It needs to label the object position, via the interested information distribution in the scene. Scale space analysis syncretizes the accuracy in the fine space and the division ability in the rough space. It can resolve the automatic or semi-automatic image segmentation for backgrounds and object regions. In view of these problems, scale space introduces entropy as its parameter. The analysis of visual information distribution could describe the essence of these phenomena. This thesis is mainly about as follows:1. Statistical characteristics between group images and individual images are significantly different. In this thesis, entropy scale space description can distinguish between backgrounds and object regions, and Harris corner detector makes further evaluation and validation.2. Quadtree is an effective hierarchy data structure to record an image, and it is also a foundation of image understanding. In this thesis, the top-down and bottom-up strategies give an attention simulation approach to locate the visual information.3. Gabor filter, which has gaussian speciality, can resolve visual information compression in adaptive image display. Meanwhile, a procedure of automatic target detection is shown by both the visual selective attention based on entropy and the object recognition using PCA (principal component analysis) method.