Efficient motion estimation schemes for medical video coding

Abstract

Motion estimation forms an important part of video compression systems but is, in general, computationally expensive. There is a growing need for motion estimation algorithms that reduce the processing while yielding good motion estimates, especially for high bit rates sources such as medical video. A common dilemma in video coding is that by speeding up the motion estimation and consequently reducing the processor cost and enabling it to work in real time, the accuracy of the estimate is adversely decreased, causing a reduced compression ratio. A trade-off thus exists between the compression ratio and the processors speed. This paper investigates several novel techniques for achieving good motion estimates at a relatively high speed. The efficiency of these schemes are compared to classical motion estimation techniques in terms of compression capabilities versus processing requirements and are found to be superior to commonly used techniques.