The Research On Medical Ultrasound Image Despeckling Method

The medical ultrasonography has a series of virtues such as non-damage andreal-time. However, owing to the backscattered of coherent echo signal, medicalultrasound images generally suffer from speckle noise, which greatly affect the imagequality. In this thesis the characteristics of speckle noise in medical ultrasonography, threemethods for speckle reduction are presented.The wavelet transform is very important in image denosing. With consideration oflimitation of despeckling method using single wavelet, this thesis firstly presents a hybriddespeckling method based on discrete wavelet transform (DWT) and dual-tree complexwavelet transform (DTCWT). This method thresholds the image by DWT and DTCWT toexploit their advantages to the full and performs well on the aspect of despeckling.Information entropy reflects the average of image information. Focused on the lowcontrast and less local information of medical ultrasound images, this thesis proposes aquantum inspired despeckling method for medical ultrasound images based on localentropy. This method combines local entropy and quantum theories to improve thetraditional dual variable shrinkage function by the dependency of wavelet coefficients andtheir parents, which can effectively reduce speckle noise on medical ultrasound images.Finally, this thesis exploits image denoising algorithms in spatial domain andtransform domain to propose a medical ultrasound image despeckling method based onbilateral filter and wavelet transform. The method first conducts quantum inspiredadaptive bilateral filtering as pre-processing, and then shrink the wavelet coefficientsaccording to Bayesian maximum a posteriori theory. The proposed method can obtainbetter despeckle and preserve detailed feature of the image.The methods proposed in this thesis are used to process the simulation speckledimages and real ultrasound images. The results show the feasibility and effectiveness ofthe methods, which can not only better despeckle real medical ultrasound images, but alsoeffectively maintain the texture details.