Despeckling Algorithm For Medical Ultrasound Images Based On Wavelet And Bilateral Filter

In medical ultrasound images, speckle noise appears as a granular structure. Due to the limitation of ultrasound imaging principle, speckle noise is inherent in medical ultrasound images, and speckle noise severely degrades the quality of ultrasound images. Speckle reduction is not only an important step in the analysis and diagnostic ultrasound images, and it is the basis for the later development of diagnostic ultrasound imaging equipment automation technology. Widespread popularity of mobile and portable ultrasound scanners also requires that ultrasound imaging instrument can provide clearer ultrasound medical images.This thesis mainly focuses on the wavelet filter and bilateral filter to suppress speckle noise. Furthermore, a new algorithm based on an improved wavelet filter and bilateral filter is proposed. According to the statistical properties of medical ultrasound image in wavelet domain, an improved wavelet threshold function is proposed. The wavelet coefficients of noise-free signal are modeled as generalized Laplace distribution. The wavelet coefficients of speckle noise are modeled as Gaussian distribution. The Bayesian maximum a posteriori estimation is applied to get a new wavelet shrinkage algorithm. The coefficients of the low frequency part in wavelet domain are filtered by the bilateral filter. Then the filtered image can be obtained by the using of the inverse wavelet transform. The comparative experiments with seven other despeckling filters are conducted. The comparative experiments shows that the proposed filter has a strong despeckling ability and it can keep the image details, such as the edge of the lesion. The main work and achievements are listed as follows:1. On the basis of a common wavelet threshold function, a new wavelet threshold function is proposed according to the medical ultrasound images in the wavelet domain statistical characteristics.2. The wavelet coefficients of noise-free signal and speckle noise are modeled as generalized Laplace distribution and Gaussian distribution, respectively A new wavelet shrinkage algorithm is got by the use of the Bayesian maximum a posteriori estimation. 3. Because there is speckle noise at the low-frequency part in wavelet domain, therefore, the low-frequency part will be filtered by bilateral filter, and the high-frequency part will be filtered by wavelet shrinkage. This is a new method to despeckle using wavelet filter.4. During the despeckling experiments of ultrasound breast image, the no-reference image quality metric (NIQE) is applied to compare the despeckling performance for the first time.5. The despeckling simulating experiment is conducted compared to seven other different filters, and two simulating methods are used to create speckle.6. In order to facilitate the proposed algorithm and other despeckling algorithm to be realized in practical applications, the graphical user interface (GUI) software of despckling is developed, which is named as "The despeckling software of ultrasound medical images v1.0".