Logarithmic number system (LNS) in MPEG encoding system

The MPEG (Moving Picture Exports Group) system is the most widely used video compression system. As more and more video clips are recorded in real time and transfered through the Internet, portable video compression cameras are desirable. Currently, most portable devices are implemented by fixed-point number arithmetic. However, the fixed-point number system has a rather small dynamic range. The Logarithmic Number System (LNS) has a large dynamic range and a low power property so that it is suitable for these portable video encoding devices.; In this dissertation, LNS arithmetic is used in some of the key parts of MPEG encoding process: DCT (Discrete Cosine Transform)/IDCT (Inverse DCT) and motion estimation. This dissertation shows LNS achieves shorter word width, less area with comparable visual and numerical results at the same precisions as fixed-point numbers.; In LNS DCT hardware, LNS can make use of the property of some fast DCT algorithms to achieve more area savings. This dissertation shows a combined LNS adder/subtractor technique to share a common table-lookup part of two LNS adder/subtractors. This technique can achieve significant area savings with the same functionalities of two LNS adder/subtractors.; The motion estimation part of MPEG encoding can also incorporate LNS arithmetic. This dissertation shows that LNS motion estimation cost function can achieve approximately the same performance as fixed-point cost functions.; With main MPEG encoding parts implemented in LNS arithmetic, a portable MPEG encoding system will be practical.