Pathological Motion Detection for Robust Missing Data Treatment

Abstract

This paper presents a new missing data detection algorithm that is robust to pathological motion (PM). PM causes clean image data to be misdiagnosed as missing data, resulting in damage to the image during restoration. The proposed algorithm uses a probabilistic framework to jointly detect PM and missing data. It builds on an existing technique of using five frames for detection instead of the standard three frames. This allows the temporally impulsive intensity profile of blotches to be distinguished from the quasiperiodic profile of PM. Another diagnostic for PM is defined on the motion fields of the five-frame window. This follows the observation that PM results in motion fields which are not smooth. A ground truth comparison with standard missing data detectors shows that the proposed algorithm dramatically reduces the number of falsely detected missing data regions. It is also shown to reduce image damage during missing data treatment.