Reducing temporal redundancy in MJPEG using Zipfian estimation techniques

Motion JPEG simply compresses each frame in a video sequence using JPEG still image compression standard. This video codec satisfies power consumption and real-time properties of embedded systems while having very low complexity. However, the downside is its inefficiency in reducing the bit-rate because there are still redundancies between encoded frames. In this paper, we focus on reducing temporal redundancies in MJPEG using a motion detection algorithm, called Zipfian estimation. The Zipfian estimation helps MJPEG extract separately the moving blocks and the stationary blocks then removes temporal redundancies by minimizing encoded bit-stream of stationary blocks and encoding only the residuals of the moving blocks between adjacent frames. Experimental results show that the proposed method can provide twice compression ratio as much as the conventional MJPEG, and an approximate quality and bit-rate as the H.264/AVC (Intra-mode). Even the proposed method uses the Zipfian estimation, it still has smaller number of operations than the conventional MJPEG if the percentage of the static scene is equal or greater than 60%. Compared with the H.264/AVC Intra mode when turning off the rate-distortion optimization, the running time of the proposed method is still only a half.