The Research On Image Denoising Using Wavelet Transform

Wavelet analysis is the local transform of time and frequency. It is now widely used in the signal processing. Along with the improvement of wavelet theory, it's application becomes more and more extensive. Donoho and Johnstone proposed a new way to denoise the noisy signal and images:"wavelet shrinkage". This paper emphasize on research of denoising images using wavelet.We proposed two method, the one is: Spatially Adaptive image denoising based on wavelet edge detection, and the other is : image denoising based on complex wavelet transform and interscale model. The first method gives a new way to estimate the variance of wavelet coefficients, it divides the wavelet coefficients into many small areas by edges detected before, and each wavelet coefficients have two states: edge point or not. To the non-edge point, it's variance is estimated by it's adjacent non-edge points in the same area. Compare to common methods, this method is more efficient and more exact,and if images aren't polluted severely, this method can get better result. Another method using complex wavelet transform and interscale model is more complex and need more time to get the final result, but it is more accurate. This paper builds a intersale models for the module of complex wavelet coefficients. The interscale model can use the dependence of complex wavelet coefficients and it's father coefficients. It's very useful to denoise images which are polluted severely, but this models can't provide clear expression,each coefficient needs a alternate compute process to get the final result. So this model need more time, but it can get a better result.