Sar Image Denoising Based On Multiscale Geometric Analysis, And Integration

High-resolution synthetic aperture radar (synthetic aperture radar, SAR) as all-weather, all-time detection and tracking investigative, and can not directly over a region mapping to the region and detection, which can be detected in standoff . The resolution of SAR is not related to the distance, which ones the infrared and optical sensors, so the resolution will not decrease as distance increases, so is the current radar SAR image classification and recognition applications such as the research focus. Based on the characteristics of SAR images, the main advantage of a new multi-scale geometric analysis tools sharp frequency localization Contourlet transformation, ie FLCT transform (Sharp Frequency Localization Contourlet Transform) (anisotropy and strong directional selection) on the SAR images were speckle reduction, based on the NSCT (Nonsubsampled Contourlet Transform) transform the multiple fusion. Its main contents are:1. Brief SAR image denoising and fusion, classification and recognition, and research status and progress. Geometric analysis of the proposed, and the research background, meaning and content.2. Analyzes the geometric analysis Contourlet transform the basic principle and on this basis developed FLCT transform and NSCT transform, multiscale geometric analysis are summarized in the image processing application.3. Analysis of the LEE filter and the lack Contourlet denoising method for multi-scale transformation, we proposed the use of KL transform domain FLCT SAR image denoising methods. Experimental results show that the method can effectively restrain the SAR image speckle noise, while the image edges and details have better ability to protect and improve the image of the peak signal to noise ratio (PSNR) and the equivalent number of looks.4. Proposed based on NSCT transformation method, the measured in the same area, the pieces focus on different SAR images combined with the local gradient and adaptive scale factor fusion rules, the scale coefficient of each sub-band integration according to different rules, and then obtained after the fusion inverse transform scale factor to be more clear image. This method is more effective than commonly used in fusion fusion.