A Framework for Very High Performance Compression of Table Tennis Video Clips

T.L. Yu (USA)


Video Compression, MPEG4, Active Contour Model, Scene Graph, Table Tennis


In this paper, we present a framework to compress video clips of Table Tennis ( Ping Pong ) with very high compression ratio for distribution or playback in embedded systems. Image processing techniques are utilized to extract the crucial parameters of the scenes in a table tennis video frame. In particular, the popular Active Contour Model is employed to track the players. The scene parameters thus obtained will be used for animation by the decoder and are referred to as scene animation parameters ( SAPs ). The SAPs of a frame are compressed using a simple DPCM model, where the differences between predicted SAPs from a previous frame and the SAPs of the current frame are calculated, transformed, quantized and entropy-encoded. The decoder carries out the usual entropy-decoding, inverse-quantization, inverse transformation, and summation to obtain the predicted SAPs of each frame. It then uses a predefined graphics model to reconstruct the scene from the SAPs. Regardless of the original frame size ( image width x height ), data rates as low as 0.3 kbps can be obtained when the video is played at 20 frames per second.

Important Links:

Go Back