Encoding Algorithms for MPEG-4 Systems and Applications
The MPEG-4 standard has just been completed last year. It is the first standard to support object-based coding. Also, error resilience is a key feature of this new standard, making it suitable for video transmission over error-prone channels. The first application of MPEG-4 will be mobile videophone terminals. As part of this project, we are developing algorithms for efficient object-based encoding and transmission. On the encoder side, this includes object segmentation algorithms, rate control for multiple objects and fast motion estimation. To assist with transmission, we consider efficient means of transcoding for bit-rate reduction and reduced resolution.
Background & Objective: The goal of this project is to develop algorithms that support an MPEG-4 object-based coding and delivery system. We have focused our research efforts on critical parts of the encoder that affect the quality and complexity, namely the rate control and fast motion estimation. In order to support object-based functionalities, we have also considered algorithms to perform the object segmentation. Finally, to assist in the delivery of MPEG-4 bitstreams over a network that is bit-rate constrained or to a user device that is limited in display size or computational power, we have considered algorithms to perform transcoding. The purpose of the transcoder is to convert the original bitstream into a new bitstream that meets constrains imposed by the network or limitations in the user device.
Technical Discussion: A rate control scheme to support the object-based functionality of MPEG-4 has been developed. The rate control for multiple video objects is unique in that shape information must also be coded along with texture and motion for each object. For object segmentation, the algorithm can be broken down into several steps. The first step estimates an initial object boundary for a sequence of images. The second step orders the space exterior to each image for effective searching. The final step minimizes an energy function that incorporates visual discontinuity, motion discontinuity and smoothness over time. To perform object-based transcoding for bit-rate reduction, an efficient dynamic programming approach has been developed. This algorithm is capable of making optimal trade-offs in spatial vs temporal quality.
Outside Collaborations: This project is done in collaboration with Princeton University.
Contacts:
Anthony Vetro
Huifang Sun
Technology Area: Audio Video Processing
Modification Date: September 12, 2007
