Visual Attention Model And Its Application In SAR Image

Studies have shown that, from the human environment in information obtainedabout80%from the visual. Visual attention is an important part of the human visualsystem. Human in the face of a complex scene, quickly shifted to the region of interest,and these regions are priority processing, the processing mechanism of human eye iscalled visual attention (visual attention). In life, the human daily access to large amountsof information, and to efficiently process the information processing, visual attention isin such environments. While the visual attention model is the simulation of humanbiological visual attention mechanism, and using the method of mathematics modeling,combined with the computer processing, thereby forming the image processing is animportant research field. Visual attention is an interdisciplinary research project,involving biology, computer vision, image processing, science and other fields, can beused for video image compression, target detection, target recognition, image retrieval,active vision etc.The main contributions of this thesis are as follows:1. In studies the of the visual attention model at home and abroad, and combining thebiological mechanism, proposed a improved visual attention model, the model throughthe extraction of image texture features and wavelet properties, generate better saliencymap, has fast calculation speed, accurate positioning, the outline is clear.2. Based on visual attention model of compressed sampling, and from the compressedsensing of visual attention models, this paper puts forward a based on visual attentionmodel can be used for the low resolution SAR image control in high resolution SARimage sampling rate so as to realize the goal of new method of compressed sensingreconstruction.3. Based on an improved visual attention model SAR airport detection method, themethod first through the visual attention generated significant chart, and then use theright amount of support machine (SVM) method to get the airport detection map. Thismethod solves the problem of the traditional method of visual attention detection resultsare not accurate problem has obvious effect, detection, contour clear advantages, can beused for the detection of SAR image airport.4. Combined with improved visual attention model this paper also presents a newmethod of automatic segmentation of SAR airport, this method first uses the visualattention model to generate significant figure, and then using the inhibition of return mechanism for automatic selection of seeds, and region growing method on SAR imageprocessing, segmentation maps obtained airport. The method is novel, use of visualattention to significant detection characteristics, the segmentation effect is good.This work was supported by the National Natural Science Foundation of China (No.60971128); Huawei innovation research project (No. IRP-2011-03-04).