Research And Application Of The Wavlet Analysis On Metal Images Denoising

The defect detection of metalwork surface has important practical significance, so the processing of metal image becomes particularly important. This paper proceeds from analyzing denoising algorithms of standard images, designs excellent denoising algorithm and uses it for metal image processing.Currently, the application of wavelet analysis for image denoising is one of the research hot spots. This paper deeply analyses wavelet transform and its application about image denoising. A new method of adaptive wavelet multi-threshold for image denoising is introduced, and the method is used in metal image denoising. The results of denoising are better than traditional methods.The main contends of this paper as follows:Firstly, the essential elements and methods are introduced, and the processing for metal images. The main contents are some general and new algorithms of image denoising. Characters of metal images are simply recommended, and so is the especial requirement and significance of metal image denoising.Secondly, the essential theory of wavelet analysis is discussed in detailed. In this part, Continuous Wavelet Transform(CWT) and Discrete Wavelet Transform(DWT) are presented, the math characteristics of wavelet basis are analyzed, and the influences and effects for practical application are briefly explained.Thirdly, the application of wavelet analysis for image denoising is introduced, three methods of wavelet denoisiong are described: modulus maximum method, spatially correlation method and threshold method. Threshold denoising is emphases in this part, and it is foundation of the adaptive multi-threshold in this paper.Fourthly, an adaptive multi-threshold denoising method is introduced, and a new algorithm of this kind is advanced, and its superiorities are proved by experiments.Lastly, the application of the new method is used in metal image. The metal image with high qualities is considered as the standard images, and the salt noises are added into the image, and then denoise it using the new method. Compared with the classical wavelet denoising methods, quantified indexes are better.