IMAGE SEQUENCE FILTERING IN QUANTUM-LIMITED NOISE WITH APPLICATIONS TO LOW-DOSE FLUOROSCOPY

Clinical angiography-the procedure of acquiring radiographic (fluoroscopic) image sequences of patients from x-ray based medical systems-has unquestionably aided cardiologists in their assessment of coronary disease. During such trials, however, literally hundreds of x-ray images are gathered, thereby putting these patients and particularly the medical staff at risk. It is desirable to lower the clinical dosages in use to abate this potential danger. With the dosage reduction, however, comes an inevitable sacrifice in image quality. In this paper, the latter problem is addressed by first modeling the noise that arises as a result of this dosage reduction. It is well-known that this noise is signal-dependent and Poisson-distributed. A model for this type of noise in image sequences is formulated and the commonly utilized noise model for single images is shown to be obtainable from the new model. We propose stochastic temporal filtering techniques to enhance clinical fluorosocopy sequences corrupted by quantum mottle. The temporal versions of these filters as developed in this paper are more suitable for filtering image sequences, as correlations along the time axis can be utilized. For these dynamic sequences, the problem of displacement field estimation is treated in conjunction with the filtering stage to ensure that the temporal correlations are taken along the direction of motion to prevent object blur.