Transforms and quantization design targeting the H.264/AVC intra prediction constraints

Authors:
Robson Dornelles;Felipe Sampaio;Daniel Palomino;Luciano Agostini
Affiliations:
Federal University of Pelotas -- UFPel, Pelotas -- RS -- Brazil;Federal University of Pelotas -- UFPel, Pelotas -- RS -- Brazil;Federal University of Pelotas -- UFPel, Pelotas -- RS -- Brazil;Federal University of Pelotas -- UFPel, Pelotas -- RS -- Brazil
Venue:
Proceedings of the 22nd Annual Symposium on Integrated Circuits and System Design: Chip on the Dunes
Year:
2009

Citing 2
Cited 2

Multiplierless and fully pipelined JPEG compression soft IP targeting FPGAs

Microprocessors & Microsystems
Overview of the H.264/AVC video coding standard

IEEE Transactions on Circuits and Systems for Video Technology

Low latency and high throughput dedicated loop of transforms and quantization focusing in the H.264/AVC intra prediction

ICIP'09 Proceedings of the 16th IEEE international conference on Image processing
Efficient hardware solution for practical intra h.264/SVC video encoder implementation

Proceedings of the 24th symposium on Integrated circuits and systems design

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper presents an architecture for a dedicated transforms and quantization loop targeting the Intra Prediction of the H.264/AVC video coding standard. The transforms and quantization loop is a bottleneck for the Intra Prediction, since the prediction process cannot start before the transforms and quantization loop finishes the reconstruction of the reference blocks. This work presents a low latency and high throughput architecture of the transforms and quantization loop, intending to reduce the impact of this loop in the Intra Prediction performance. The architecture was described in VHDL and synthesized to Altera Stratix II FPGA and to the TSMC 0.18μm CMOS technology. The architecture reaches an operation frequency of 129.3 MHz when mapped to standardcells, and takes 4 cycles to process an I4MB block, 24 cycles to process an I16MB block, and 22 cycles to process a chroma block. Those are the best results when compared with all related works considering the Intra Prediction constraints.