Selective Visual Attention Mechanism And Its Application In Image Processing

Synthetic Aperture Radar (SAR), as the active radar, was proposed as early as in the1950s. Because SAR can work in nearly all weather conditions, day and night, it has been widely used in many fields.In recent years, with the development of SAR imaging technology, more and more high-resolution SAR images appear, and more and more attention are being paid in the technology of SAR image target detection and recognition. Since there is much difference in imaging mechanism between SAR imaging and optical imaging, and the SAR image is also affected by coherent speckle noise, many natural image processing algorithms can not be directly applied to SAR images. Additionally, SAR images are usually massive data, so it also poses a challenge to the efficiency of image processing for the computer.However, in SAR image target detection, we note that the targets to be detected are usually a small part of the whole image, so it is not necessary to process all parts of the whole image equally. If human visual attention mechanism is introduced, mimicking the selectivity and activity of human vision, finding the regions of interest as the significant areas quickly and ignoring or giving up the insignificant areas, it would greatly improve the efficiency of image processing. This thesis studies the visual attention model and proposed a method of extracting visual significant areas, and then introduces the visual attention model to the target detection of SAR images, and completes the relevant experiments. Experimental results show that introducing the visual attention model to the target detection in SAR images not only improves the efficiency of the algorithms, but has better detection results as well.The main contributions of this thesis are as follows:1. Proposed a bottom-up image salient regions detection method. First, a bottom-up selective attention model is constructed based on the theory of psychological andphysiological studying of viaual attention mechanism. Thdn, the model is introducted into image information processing, and completed image salient regions detection.2. Proposed a top-down SAR image waters segmentation methos. First, a top-down selection visual attention model is constructed based on the theory of psychological andphysiological. Then, the waters saliency map and seed points are obtained based on the constructed model. At last, according to the visual conspicuity of waters, all of the waters are extracted by region-growing method.3. Proposed a SAR image ship detection algorithm based on the selective visual attention mechanicsm. The proposed algorithm combines the bottom-up and the top-down two selective visual attention mechanisms. First, the selective visual attention mechanism is selected automatically according to the gray value distribution of SAR image. Then, construct the selected visual ateention model which is used to complete SAR image ship detection.