MPEG IMAGE COMPRESSION AND ATM NETWORKS

ATM TRANSPORT AND CELL-LOSS CONCEALMENT TECHNIQUES FOR MPEG VIDEO

By kulanthai Chandrarajh.
Information System Engineering(2nd Year)
Imperial College of Science , Technology & Medicine
Department of Computing
180, Queensgate
London SW7 2AZ,
United Kingdom

ABSTRACT

ATM (asynchronous transfer mode) transport and cell-loss concealment techniques for MPEG (Moving pictures experts group) video are discussed in this paper. Robust video compression and transport approaches are essential for operation over emerging ATM-based broadband ISDN and "gigabit" networks characterized by packet(cell) losses due to conjestion. A specific cell-based MPEG video transport approach applicable to a variety of media is described for both one-tier and two-tier transmission alternatives. Decorder cell-loss concealment techniques supported by this transport formats are introduced, and typical design issue are discussed.

1. INTRODUCTION

This paper presents a review of ATM ( asynchronous transfer mode) transport and cell-loss concealment techniques for MPEG* video. Transport and concealment methods for compressed video are of increasing importance in a variety of communicative and distributive TV and multimedia applications. Robust video compression and transport approaches are essential for operations over emerging ATM-based broadband ISDN networks in which occasional packet(cell) losses can be experienced.

In the "MPEG++" video delivery approach described in this paper, transmission of MPEG compressed data is supported by an appropriate packet (cell) transport protocol. This "ATM" transport format reliably detects packet loss/error, helps identify the picture area in which the compressed information was lost, and provides for resynchronization of video decoding. The displayed picture quality can be maintained at an acceptable level using a process known as "error concealment" by which the decoder subtitutes pridicted information in pictures areas identified as damaged.

The rest of this paper is organized as follows. Sec. 2 presents an ATM/cell based format for transport of MPEG compressed video. In Sec. 3, general principles of error concealment algorithms are described.

2. MPEG TRANSPORT

2.1 Data Transport Format:

The MPEG transport layer is a packet oriented approach to reliable video delivery, consisting of two distinct sublayers: "data link level" and "adaptation level". The data link sublayer supports features such as service multiplexing and error detection. This layer formats the MPEG bit-streams into a series of fixed length L byte cells, each contaning (L-8) data bytes.

2.2 Data-link sublayer:

The 1-byte data-link header contains generic transport information such as cell priority, service ID to support multiplexing, and a cell sequence number for positive identification of error events in each service stream. In this implementation the nominal 1-byte service header format provides an address space of 7 primary video , data or control services per channel.

2.3 Adaptation sublayer:

This has been designed to permit rapid decoder resynchronization after error events that result in the loss of one or more cells. Adaptation headers contain information fields which aid error recovery at the video decoder. For MPEG encoded video, these fields include frame type indicators, slice IDs etc. Interpretation of these header fields enable the decoder to locate slices received in error, and then resume decoding soon after error events.

2.4 MPEG Data Prioritization:

In considering the data Prioritization function, it is observed that MPEG produces a bit-stream consisting of variable length codewords conveying information about different picture attributes such as headers, motion vectors, run-lengths, etc. The impact of losing each of these codeword types in the presence of channel errors would have more serious effect on the received picture quality. Accordingly, separation of MPEG data into higher prioriy(HP) and standard priority(SP) streams may carried out using an adaptive Prioritization algorithm .

3. CONCEALMENT

Receiver error concealment is intended to reduce the impact of lost video data by exploiting available redundancy in the decoded picture. Once the image regions to be concealed are identified, a combination of temporal and spatial replacement techniques may be applied to fill in the lost picture elements. In MPEG compression, video frames to be coded are formatted into a group of pictures(GOP) consisting of a sequence intra-coded (I), predictive-coded (P) and bidirectionally predictive-coded (B) frames. Intra-coded (P) frames are encoded spatially and used as anchor frames for motion-compensated forward prediction of P frames and forward/backward prediction of B frames. This structure of MPEG implies that if an error occurs within I-frame data, it will propagate through all frames in the GOP. Similarly, an error in a P-frame will affect the related P and B frames, while B-frame errors will be isolated. Therefore, it is desirable to develop error concealment technques which prevent error propagation from I-frames and, consequently , to improve the quality of reconstructed pictures.

Two methods are available for error concealment at the decoder.

Temporal replacement attempts to estimate lost picture region from previusly received anchor frames, using the closest available motion information. For P and B frames, if motion information is available, temporal replacement is always the preferred method. However, for I-frame there is no motion information, so that significant differences might exist between the current intra-coded frame and previusly decoded frame. In this case, temporal temporal replacement will produce large distortion. Accordingly, I-frame concealmnt is based on spatial interpolation approach of lost picture elements from adjacent pixels(blocks) in the same frame. In spatial interpolation, intra-frame redundancy between blocks is exploited, while a potential problem of blurring remains. To address this problem, an adaptive error concealment technique has been developed and evaluated. The decision regarding which concealment method should be used is based on simply obtained measures of image activity from the neighoring (top and bottom) macroblocks. If the local motion is smaller than spatial detail, the corrupted blocks belong to the class on which temporal replacement is applied; when local motion is greater than local spatial detail, spatial interpolation is used.

3.1 Recent Progress in error Concealment:

Improvements in Spatial interpolation algorithms have been proposed. In these studies, additional smoothness criteria and/or directional filtering are used for estimating the picture area to be replaced. The new algorithms utilize information from a large local neighborhood of sorrounding pixels and perform interpolation to restore the missing block. A second approach under consideration is based on transmission of side information, i.e. I-frame motion vectors to be used as an aid for error concealment. A third approach for error concealment of MPEG video is based on the scalability feature of MPEG-2. Hierarchial transmission provides more possibilities for error concealment from the scaled data, provided corresponding two-tier transmission media are available.

4. CONCLUDING REMARKS

In this paper, principles for ATM transport and error concealment of MPEG video have been discussed. A specific "MPEG++" implementation for robust digital TV/HDTV delivery via one- or two-tier transmission media has been described. It has been demonstrated that robust transmission of compressed MPEG video can be achieved with a combination of appropriate ATM-type transport and decoder processing techniqes.

REFERENCES

  1. Advance Television Research Consotium, "Advanced Digital Television: Prototype Description", FCC WPI certification Document,feb 1992.
  2. H. Sun and W. Kwok, "Adaptive Concealment for Block-based compression video".
  3. W. Kwok and H .Sun, "Multi-directional Interpolation for Spatial Error Concealment," to be submitted.
  4. H .Sun and j.Zdepski ,"Adaptive Error Concealment Algorithm or MPEG compressed video", SPIE Proc. Visual Comm. and Image Processing 92, Vol. 1818, Nov. 18-20, 1992, pp. 814-824.
  5. F. Kishino, K. Manabe, Y. Hayashi, and H. Yasuda, "Variable Bit-rate Coding of Video Signals for ATM Networks," IEEE J.Selected Areas in Comm., Vol 7., No. 5 ,june 1989.
  6. MPEG video Compression, Author "sab"
  7. Is ATM the future of all global communications? By "arad".
RETURN HOME