Fast Algorithm and Architecture Design for MPEG-4 Shape Coding

Introduction

MPEG-4 is a new international standard for multimedia communication. It provides a set of tools for object-based coding of natural and synthetic videos/audios. MPEG-4 also enables content-based functionalities by introducing the concept of video object plane (VOP), and such a content-based representation is a key to enable interactivity with objects for a variety of multimedia applications. The VOP is composed of texture components (YUV) and an alpha component, which is shown in Fig. 1. The texture component contains the colorific information of video object, and the alpha component contains information to identify the pixels, which are inside an object (opaque) and the pixels, which are outside the object (transparent). MPEG-4 supports a content-based representation by allowing the coding of the alpha component along with the objects¡¦ texture and motion information. Therefore, MPEG-4 shape coding becomes the key technology for supporting the content-based video coding.

Fig. 1 The concept of object based video coding

Feture

n Shape encoder takes 20% of total computational complexity of MPEG-4 encoder

n Binary motion estimation (BME) takes nearly 90% of total computational complexity of shape encoder

Result

Fig. 2 Performance comparisons of SAD and WSAD using full search algorithm for the test bream sequence.

Fig. 3 Performance comparisons of SAD and WSAD using full search algorithm for the test news sequence.

Fig. 4 Performance comparisons of SAD and WSAD using full search algorithm for the test foreman sequence.

Fig. 5 Performance comparisons of SAD and WSAD using full search algorithm for the test children sequence.

Fig. 6 Performance comparisons of various search algorithms for the test bream sequence.

Fig. 7 Performance comparisons of various search algorithms for the test news sequence.

Fig. 8 Performance comparisons of various search algorithms for the test foreman sequence.

Fig. 9 Performance comparisons of various search algorithms for the test children sequence.

Table I Total search positions for various search algorithms.

Sequence	Full Search	Ref. [6]		Ref. [7]		Proposed (BS)		Proposed (DBS)
	SP	SP	%	SP	%	SP	%	SP	%
Bream	10533897	6506028	61.76	1593173	15.12	368172	3.50	67164	0.64
News	791703	431816	54.54	7697	0.97	27143	3.43	2281	0.29
Foreman	9100773	5096208	56.00	1573951	17.29	287818	3.16	57311	0.63
Children	13323915	8452318	63.43	1806706	13.56	534021	4.01	112484	0.84

Table II Average bit-rate for various search algorithms.

Sequence	Full Search	Ref. [6]		Ref. [7]		Proposed (BS)		Proposed (DBS)
	Bits/VOP	Bits/VOP	%	Bits/VOP	%	Bits/VOP	%	Bits/VOP	%
Bream	1599.80	1599.78	100.00	1616.96	101.07	1591.84	99.50	1600.41	100.04
News	890.22	890.22	100.00	880.41	98.90	892.20	100.22	891.55	100.15
Foreman	1186.94	1186.97	100.00	1187.61	100.06	1180.05	99.42	356279	100.06
Children	2056.03	2056.01	100.00	2062.81	100.33	2074.69	100.91	2077.41	101.04

Table III Runtime simulation results of various search algorithms.

Sequence	Full Search	Ref. [6]		Ref. [7]		Proposed (BS)		Proposed (DBS)
	ms	ms	%	ms	%	ms	%	ms	%
Bream	16606.09	21330.50	128.45	1479.46	8.91	2318.21	13.96	362.15	2.18
News	1579.97	1878.64	118.90	21.45	1.36	239.42	15.15	10.85	0.69
Foreman	20349.37	25637.73	125.99	3226.78	15.86	2792.07	13.72	615.20	3.02
Children	21572.47	28251.50	130.96	3196.80	14.82	3283.70	15.22	925.99	4.29

Fig. 10 Chip layout of MPEG-4 shape encoder.

Table IV Chip Features

Made by ¼ï¦t¨k