DocumentCode
3547853
Title
A hardware-based predictive motion estimation algorithm
Author
Hamalainen, Saku ; Koskinen, Lauri ; Halonen, Kari
Author_Institution
Electron. Circuit Design Lab., Helsinki Univ. of Technol., Espoo, Finland
fYear
2005
fDate
23-26 May 2005
Firstpage
6114
Abstract
For low-power motion estimation hardware solutions, advanced hardware aimed algorithms and architectures are a necessity. A hardware solution for the modern emerging standard H.264 is investigated. Design trade-offs, including search patterns, predictors, and memory architecture, have been made with power consumption and hardware complexity in mind. The memory architecture has been optimised for the variable block sizes used in H.264 motion estimation. All predictors are refined with the small search pattern. An average PSNR of -0.0438 dB compared to the full-search algorithm was simulated. A 2.793 mm2 layout, using 0.18 μm CMOS technology, of the design running at 7.61 MHz is presented.
Keywords
CMOS digital integrated circuits; hardware-software codesign; integrated circuit design; memory architecture; motion estimation; power consumption; prediction theory; search problems; video coding; 0.18 micron; 7.61 MHz; CMOS technology; H.264; hardware complexity; hardware-based predictive motion estimation algorithm; low-power hardware solutions; memory architecture; power consumption; predictors; search patterns; Algorithm design and analysis; CMOS technology; Clocks; Energy consumption; Hardware; Memory architecture; Motion estimation; Partitioning algorithms; Prediction algorithms; Streaming media;
fLanguage
English
Publisher
ieee
Conference_Titel
Circuits and Systems, 2005. ISCAS 2005. IEEE International Symposium on
Print_ISBN
0-7803-8834-8
Type
conf
DOI
10.1109/ISCAS.2005.1466035
Filename
1466035
Link To Document