The Study On Improved Wavelet Threshold Denoising Method

Wavelet threshold denoising method is a simple and effective denoisingmethod, but to some extent, there are some defects on the traditional waveletthreshold method in practical application. Therefore, it is quite necessary to dothe further research on the wavelet threshold method. The traditional universalthreshold is fixed on each scale and does not have the adaptive capacity, whichhas a great impact on the denoising effect. The hard threshold function isdiscontinuous at the threshold points, for the soft threshold function, although itis continuous at the threshold points, there exists the constant deviation betweenthe soft threshold function and the real signal. The wavelet transform is effectivefor the detection of point singularity, but it can not deal with the line singularityof the image effectively. However, this drawback of wavelet transform can becovered by the ridgelet transform. The line singularity can be detected well bythe ridgelet transform. To achieve the purpose of removing the noise effectively,the wavelet threshold method is improved in this paper. The specific work is asfollows:Firstly, the universal threshold is improved in this paper so that it can bechanged with the decomposition level. Compared with universal threshold, theimproved threshold is flexible and accords to the distribution of noise in eachlayer. Then the threshold function is improved in this paper, a parameter is addedto the improved threshold function. By adjusting the parameter, the improvedthreshold function is between the two threshold functions, it can overcomediscontinuity of the hard threshold function and reduce the deviation between thesoft threshold function and the real signal. The improved threshold function iscontinuous and differentiable. In this way, it is rather convenient to do somesubsequent processing.Secondly, the wavelet threshold method can not deal with the line singularity effectively, but the ridgelet transform could make up this shortcoming.The images have great "wrap-around" effect, after ridgelet method has been usedto remove noise, so the traditional ridgelet denoising method is improved in thispaper. First of all, noisy image is divided into image blocks which are the samesize and non-overlapping square blocks, and the ridgelet transform is used toremove the noise of each block. Then, all of the blocks are put together to be acomplete image; next, wiener filter is used to reduce the "wrap-around" effect ofthe image.Finally, this paper makes fusion with the two images which obtained by theimproved wavelet threshold denoising method and block ridgelet denoisingmethod, in this way, the fused image shares the advantages of the two methods.The experimental results show that the improved method can remove the imagenoise more effectively and obtain a better denoising effect compared with thecommonly-used mean filter, median filter and ridgelet denoising method.