Distance dependent thresholding search for fast motion estimation in real world video coding application

This paper presents a distance dependent thresholding search (DTS) block motion estimation algorithm that employs the novel concept of distance dependent thresholds. The key feature of this algorithm is its flexibility, trading-off quality and complexity with threshold variation. Whereas the performance of the existing algorithms is fixed in terms of prediction quality as well as complexity, DTS can be used as full search (FS), where high quality video entertainments require motion estimation with small prediction error, as well as fast motion estimation such as three-step search (TSS), new-three-step search (NTSS) etc. while real-time video applications, such as the speed-oriented video conferencing require fast motion estimation with sacrificing quality. Experimental results show that this DTS algorithm also achieves better peak signal-to-noise ratio (PSNR), as well as lower search times in comparison to both the TSS and NTSS algorithms.