Speckle denoising method for ultrasonic image

1. introduction

Medical ultrasound image segmentation and recognition is an important means to analyze pathology , One of the most important factors that affect the quality of ultrasonic image is speckle noise , This seriously affects the segmentation of ultrasonic images , Therefore, speckle denoising is the key preprocessing process of medical ultrasonic image segmentation and recognition .

The main algorithms of ultrasonic image denoising include ：

• Spatial denoising algorithm
• Wavelet domain denoising algorithm
• Anisotropic diffusion denoising algorithm

2. Local statistical filtering algorithm in spatial domain

The local statistical filtering method is based on certain estimation criteria , Therefore, different estimation criteria such as the minimum mean square error criterion (MMSE)、 Maximum likelihood method (ML) Maximum a posteriori probability (MAP) And so on , Typical algorithms are Lee Filtering algorithm and Kuan Filtering algorithm .

3. Image wavelet transform denoising method

Classical wavelet transform image denoising algorithms are divided into three categories ： Modulus maxima method based on wavelet transform 、 Correlation denoising method 、 Wavelet threshold denoising method .

4. Anisotropic diffusion filtering algorithm

Compared with some traditional spatial filtering techniques , The advantage of anisotropic diffusion is that it can remove noise at the same time , Retain or even enhance the edge information in the image . Typical algorithms include PM Model 、SRAD Model 、NCD Model .

Summary of the above algorithms ：

• The spatial speckle suppression algorithm has a small amount of computation , Fast , But its filtering effect is closely related to the window size , Unable to enhance the edges and details of the image , The filtering in the edge region is isotropic .
• Anisotropic diffusion filtering and multi-scale filtering have excellent performance .

5. be based on MAP Ultrasonic image decomposition and denoising algorithm

Ultrasonic image segmentation can be regarded as an image decomposition problem , The goal is to decompose the speckle contaminated image into real image and speckle noise image , It is considered that speckle noise contains useful information , Used in subsequent image segmentation .

reference

1. A review of speckle denoising methods for medical ultrasonic images
2. be based on MAP Ultrasonic image decomposition and denoising algorithm