Research For Bilateral Filtering Theory And Its Application In Remote Sensing Image Processing

The bilateral filtering combines the Gaussian filter with the neighborhood filtering, it doesn’t only take the distance between the center pixel and its neighboring pixels into account, but also the degree of grayscale value’s similarity between them, which makes the bilateral filtering effectively smooth out noise while preserve edge information of images well. So it has been the researching focus of many scholars, having high research value and application prospect. It has been widely applied in medical image processing, video coding, tone mapping of high dynamic range images, optical flow estimation, super-resolution reconstruction of images, image fusion etc.However, when the intensity of noise in images is large, de-noising and edge-preserving performances of bilateral filtering remain to be further improved. In view of this, the gradient bilateral filtering is proposed against existing problemes and deficiencies of the traditional bilateral filtering in the paper, the general configuration of its filtering parameters is completed through experiments and experiential learning methods. In addition, for the Contourlet transform and nonsubsampled Contourlet transform have excellent characteristics of capturing image detailes such as edges and textures, and combining with the edge-preserving and smoothing characteristics of bilateral filtering, bilateral filters based on the Contourlet transform and nonsubsampled Contourlet transform are respectively proposed. Experimental results demonstrate the effectiveness of the proposed denoising algorithms.For the research on application of bilateral filtering in multi-source remote sensing image fusion, this paper focuses on the research of the bilateral filter in the application of the fusions of infrared&visible images and multi-spectral&panchromatic images. For the fusion of infrared and visible images, since traditional low-frequency coefficient fusion rules don’t consider image edge feature, and can’t improve the clarity of fused images, a low-frequency coefficients’adaptive weighted fusion rule based on regional average gradient is proposed by calculating the low frequency coefficients’local average gradients of two images to be fused in the paper, and Experiment verifies its fusion effectiveness. In addition, the proposed and improved bilateral filtering methods are used in this kind of images fusion, and compared and analyzed with several other common fusion methods through experiments. For the fusion of multi-spectral and panchromatic images, since the RGB color system does not meet people perception of color, and the most consistent system with human perception of color is IHS, a fusion framework based on IHS and bilateral filter is proposed combining with IHS transform in the paper, the proposed fusion framework obtains better visiual effect of fused images.In view of the research on application of the bilateral filtering in coherent speckle suppression in SAR images, since when the bilateral filter is used to suppress speckles in SAR images, its despeckling performance and running speed need to be improved, a despeckling method based on the template bilateral filtering for SAR images is proposed in the paper, and the estimation of its optimal despeckling parameters is achieved through experiments. Besides, since the problemes and deficiencies exist when the emplate bilateral filtering uses a fixed spatial parameter to remove speckles, the paper further improves it, the spatial parameter is adaptively selected according to local homogeneity in SAR images, so an improved template bilateral filtering despeckling method is proposed. The experimental results demonstrate the effectiveness of the proposed despeckling methods.