Reduction of Power Consumption in Video Communication based on Low Frame Rate Transmission and Decoder Frame Interpolation

Abstract

In order to reduce the power consumption of an endoscope capsule transferring still-images over a wireless channel from inside human intestines to on-body receivers, the transmitted frame rate can be reduced in favor of generating the frames at the receiver side. In this thesis, motion compensated frame interpolation methods are studied in order generate these frames. The study consists of three parts. The first two parts study motion estimation and motion compensation techniques when applied to synthesized artificial videos. In the last part, the best performing techniques are applied to real endoscope videos in order to determine the best performing algorithms.The motion estimation methods studied are unidirectional and bidirectional block-matching motion estimation (BME), and motion estimation based on optical flow fields. Applying motion compensation prove that bidirectional BME most likely produce the best result as it, among other things, does not leave unvalued pixels in the interpolated frame as the other methods do. Additionally both a nonadaptive and an adaptive vector median filter (VMF) are applied to the motion fields in order to remove erroneous vectors. Because of the expected low complexity of the motion fields in the endoscope videos, the non-adaptive VMF is found to produce the best result. Overlapped block motion compensation (OBMC) is additionally tested and proven to give great contributions to the motion compensation. Based partly on the results from a subjective survey, the proposed frame interpolation algorithm gives an acceptable result when applied to endoscope videos. A few artifacts still appear in some type of videos, but possible improvements are suggested in order to increase the performance.