Remote Sensing Image De-noising Method Based On Partial Differential Equation

The quality of remote sensing image is always degraded by noises duringacquisition and transmission, which make it difficult to identify and analysis image.So it is much necessary to effectively remove these noises before using it. In recentyears, nonlinear diffusion model based image de-noising has received considerableattention in the field of remote sensing image application since the model can wellpreserve the edge, and the method of de-noising has become a new image processingtools after wavelet.Firstly, The thesis simply analyzes of the remote sensing image and its noise.Secondly, The thesis introduces the development of partial differential equationde-noising, and their de-noising characteristics and focus on introducing the P-Mequation. Finally, the paper proposes three de-noising models based on PDE asfollows:(1)A mixed PDE model remote sensing image denoise algorithm based onregion segmentation, combined respective advantage of pure anisotropic diffusionmodel and four order PDE model. This algorithm overcomes disadvantages of smoothregion excessive diffusion and staircase effect in traditional pure anisotropic diffusionmodel, and edge information excessive loss in four order PDE model.(2) A remotesensing image denoise algorithm based on wavelet fractal and differential equation,combines each advantage in image denoise. Wavelet basis is also optimized by fractalconstant.(3)A remote sensing image denoise algorithm based on pure anisotropicdiffusion model under NSCT transform and Butterworth high-pass filter, also withseparate advantages. Our models can not only use to efficiently remove noise inremote sensing image, but also simultaneously retain detail information, such as edgeand texture. Experimental results suggest the effectiveness and stability of theproposed model.