Hardware-Oriented Block Matching with Single Precision Floating Point Division Newton-Raphson Algorithm for Fast Motion Estimation in Wire Less Networks

Abstract

In the past few decades, we have been running towards the Digital Age. Telephony networks, fiber optic networks, Internet, satellites, third-generation (3G) wireless networks, these advanced transmission technologies and network infrastructure ensure digital information can flood into every corner around us. At the same time, the traditional media format has been gradually replaced by today’s digital multimedia formats. These new formats and technologies shorten the distance between people, and improve the quality of lives. In this paper we have presented a Fast Motion Estimation technique known as Quadrant based search algorithm with parallel pipeline divider (QSPD) is constructed using Newton-Raphson Algorithm. Motion vector of the video sequence is estimated with Stationary Block Prediction (SBP). Stationary Block Prediction technique is used to find the block with zero motion. Implementation results of our proposed algorithm are compared with previous methods such as ARPS and Diamond search algorithm. PSNR value and motion estimation time of our proposed algorithm are better than that of previous method. In my research work, we concentrate on extending the block matching algorithms (BMA) from single reference frame to multiple reference frames. BMAs are selected due to their successes in the past video coding standards and the simplicity of regular data structure that is favorable for both hardware and software implementations. Analysis of motion vector distribution in multiple reference frames. Analysis the searching pattern of some out-standing BMAs and Developed a novel searching strategy for multiple reference frames based on the analyzed results. A novel approach to develop the fast motion estimation algorithm results are analyzed in Xilinx ISE and implemented in Spartan low power FPGA-6.