|M.Sc Student||Marom Ran|
|Subject||A Feature Dependent Approach for Improving Speckle Noise|
Reduction and Side Lobes Suppression in
|Department||Department of Biomedical Engineering||Supervisors||Professor Emeritus Dan Adam|
|Dr. Zvi Friedman|
|Full Thesis text|
Medical ultrasound images are severely degraded by speckle noise and side lobes artifacts, both of which result from the physics of ultrasound imaging. These two sources of noise are commonly dealt with by methods that are efficient - but have their own shortcomings: speckle noise reduction is commonly treated by frequency compounding, which has a drawback since it degrades the axial resolution. Side lobes suppression, on the other hand, is usually treated by apodization, which has its own drawback as it degrades the lateral resolution. In the present study, an improvement of these methods was proposed, by using a feature dependent approach which applies different signal and image processing tools to different sections of the image, according to the objects detected there. Frequency compounding was applied to selected regions in the image and avoided from others, depending on the identification of features for which preserving the axial resolution was of high importance. Such features are point reflectors, lesions boundaries and cysts boundaries. Side lobes suppression was applied only at specific regions in which side lobes artifacts were identified.
The proposed methods were evaluated by experiments using real ultrasound data of phantom scans.
Feature dependent speckle noise reduction was shown to be significantly advantageous when compared to standard frequency compounding, as it was capable of smoothing speckles in the image while preserving resolution and contrast of strong reflectors and edges of cysts and lesions.
Feature dependent side lobes suppression was shown to be significantly efficient when suppressing side lobes of point reflectors, producing side lobes levels which were 5.8 dB lower than those obtained by ordinary apodization, while preserving the width of the main lobe. Thus, the proposed approach has the potential of significantly improving ultrasound image quality, while specifically enhancing the important features in the image.