Medical Ultrasound Image Denoising Method Based On Wavelet Transform

Recently, with the improvement of wavelet theory, wavelet analysis has penetrated into many fields. Image de-noising in wavelet domain is one of the focus methods. Nowadays, there have been many new image de-noising algorithms based on wavelet. In this thesis, considering the characteristics of noises in ultrasonic images, we propose a new ultrasonic image de-noising algorithm based on the nonparametric adaptive estimation.Due to the complexity of noise in ultrasonic images, it raises a higher request for de-noising methods. There are two demands in ultrasonic image de-noising. First, the noise needs to be smoothed out; and in the mean time, characteristic details in the original images should be preserved as much as possible, since they constitute significant evidence in the medical diagnosis of a human organ. To meet the above demands, we have done in-depth studies on two aspects: one is studying the de-noising methods of ultrasonic images in stationary wavelet domain, and the other is confirming the wavelet threshold based on nonparametric adaptive estimation.The stationary wavelet transform is a non-orthogonal wavelet transform derived from the orthogonal wavelet transform, the difference between them is that there is no subsampling in stationary wavelet transform, every layer is of the same size as the original image. For orthogonal wavelet transform, since it implements subsampling in each decomposing, it does not meet the shifting invariance in wavelet domain, and thus after processing by threshold may cause instable phenomena, such as pseudo peaks, etc. It goes against the accurate reconstruction of edges. While the stationary wavelet transform has overcome this weakness of orthogonal wavelet transform, it realizes stability through preserving all coefficients on each scale and sacrificing the orthogonality. Also, the stationary wavelet transform can achieve quality edge characteristics in the reconstructed image.