Filtering Ultrasound Images Based on Morphological Operations

Abstract

A review of the methods of filtering ultrasound images is carried out, their advantages and disadvantages are given. It is proposed to use morphological transformation procedures for filtering. It is shown that morphological processing has several advantages, in particular, simplicity of implementation. The numerical image processing method is based on non-linear transformations of their shape. Various image processing modes based on arbitrary-shaped structural elements are used. Morphological processing allows to perform almost all operations processing black and white images. A large number of halftone operations can be performed. Filtration is checked on test images of previously noisy regular geometric shapes, as well as on real ultrasound images of the hip joint. To blur and generate noise, the standard functions of the Image Processing toolbox of MATLAB are used. The simulation was carried out in the NI Vision Assistant package. To determine the dependence of the parameterization error on the size of the object and the noise level, standard test images of the correct geometric figures were used (squares 15´15, 25´25, 35´35, 45´45, 55´55). For the most approximate reproduction of the features of a real image of ultrasound diagnostics, the reference image was blurred, after which an artificially generated speckle noise with a standard deviation of 0.25 was superimposed on it. When comparing the parameters of a quantitative assessment of the quality of the filters, it was concluded that filtering based on morphological operations according to the PSNR and MSE, √MSE indicators shows better results, and also allows to get a more informative image.