Application Of Wavelet Statistical Methods In Remote Sensing Image Filtering

As time progresses and science and technology advance,remote sensing technology has become increasingly vital in enhancing people’s living standards.Specifically,remote sensing images play a prominent role in providing people with intuitive information about the world around them.However,remote sensing images are often subject to various forms of noise during acquisition and transmission,leading to a decline in their quality.To obtain clear and high-quality remote sensing images,it is essential to apply filters that effectively eliminate noise.The filtering algorithm based on the wavelet transform combined with statistical methods is widely used,however,the wavelet basis function in the wavelet transform lacks translation invariance and cannot describe the hyperplane high-dimensional image well,while the remote sensing image has high-dimensional characteristics.Therefore,this paper considers the filtering method of multi-directional wavelet transform for remote sensing images,which can describe the hyperplane high-dimensional signal well and has better sparsity in representing the image edge information.Therefore,this paper proposes two improved algorithms based on the filtering algorithm of wavelet transform and goodness-of-fit test to improve the performance of image noise removal.The main innovations of the paper include the following two points.1.To tackle the issue of Gaussian noise removal from visible remote sensing images,and to address the limitations of existing image filtering algorithms based on wavelet transform and goodness-of-fit test,this paper proposes a novel remote sensing image filtering algorithm that leverages multi-directional wavelets and goodness-of-fit test.Firstly,the remote sensing image undergoes multi-directional wavelet decomposition,and the resulting decomposition coefficients are normalized.Next,the GOF test is applied to the coefficients locally,and the real signal coefficients are obtained after local testing.These coefficients are then inverse normalized,and the resulting coefficients are used to reconstruct a noise-free remote sensing image.The improved algorithm is experimentally evaluated,and the results indicate that the PSNR and SSIM values are significantly improved with the proposed filtering algorithm,and provides better subjective visual quality.2.The current filtering algorithm for grayscale remote sensing images using multi-directional wavelet and goodness-of-fit test has limitations,as most remote sensing images are RGB color images.In the RGB space,the strong correlation between channels can result in a poor filtering effect when any channel changes.To overcome this limitation,this paper proposes a color remote sensing image filtering algorithm that incorporates multi-directional wavelet and spatial conversion.The algorithm converts the RGB color space images into OPP color space and YCr Cb color space using color space conversion methods.These color spaces have weak correlations between channels,which is advantageous for the image filtering process.The remote sensing image is then filtered using the improved GOF-Curvelet filtering algorithm,and the reconstructed image is inverted in the color space.Finally,the RGB remote sensing image is output.Experimental results indicate that the proposed algorithm achieves better denoising performance in terms of both numerical results and subjective vision when compared with other denoising algorithms.