Frame Interpolation of Ultrasound IQ Data using Dictionary Learning and SFA Method

Frame interpolation holds the potential to overcome the physical limitations inherent in the frame rate of ultrasound imaging. We endeavor to incorporate phase information by extracting pertinent features from IQ data to decompose the frame sequence into speckle motion and target motion, and rebuilding the interpolation frames. First, the IQ frames are represented using a dictionary learning method to extract phase and speckle information from complex IQ data. And we employ a Slow Feature Analysis (SFA) method to distinguish various changes within the representation. following the decomposition, the frame sequence is segmented into separate sequences. Second, these sequences are then recovered by utilizing the dictionary, and we apply traditional optical flow methods to extract motion information from each sequence individually. Finally, we combine the three sequences to generate the interpolation frames. The proposed method achieved more fluent results than the conventional methods.