Research On Infrared Image Denoising Algorithm Based On Wavelet Transform

In recent years, the technology of infrared imaging are widely applied in manyfields. Including, night vision imaging，guidance, tracking, medical and so on. However,much noise, the low contrast and blurred edge is a serious problem to the infrared imagesbecause of the inherent character of infrared detectors. In order to overcome thesedrawbacks and improve the image quality the technique of wavelet analysis is introduced.Considering the special features of the wavelet transform, and it’s ability of decomposingthe planar signals to different scales, it becomes very proper to image analysis. After thewavelet reconstruction, the visual quality of the processed images can be improvedeffectively and obtain the information we are needed.The research is mainly focused on the infrared image de-noising based wavelettransform. First, some basic theories of wavelet transform are introduced. Thecharacteristics and selection criteria of the wavelet bases is analyzed. The noisedistribution characteristic and evaluation standards of infrared image are introduced. Thispaper introduces several classic wavelet transform denoising method. And these methodare analyzed and compared. Several key problems on wavelet threshold de-noisingmethod are discussed in detail. A new threshold function de-noising methods areproposed based on LMMSE guidelines.In many kinds of method of infrared image denoising based on wavelet transform,wavelet threshold shrinkage is widely used. Traditional denoising method are frequentlyapplied to practical and have obtained a good effect. Discontinuity of thehard-thresholding function results in Pseudo-Gibbs phenomenon of the reconstructedsignal. Soft-thresholding function has good continuity but a constant deviation of theestimated value from the actual value confines its application. According to the noisedistribution characteristic of infrared image and characteristic of wavelet bases andLMMSE standards, a new thresholding function is proposed based on waveletmulti-resolution analysis and LMMSE standards. The method is convenient to treatmathematically. At last, the theoretical basis to Matlab as a platform for denoising. The experiments results show that the new threshold function denoising better PSNR thantraditional method. At the same time, the new method can also preserve more imagedetails, so it is an effective method of infrared image denoising.