Optimizing video compression using high performance algorithms and techniques

Speaker: Vikram Kadam,

Affiliation: PEDAL Lab, School of Systems Engineering, University of Reading.

Date/Time: 4th February 2009, 13:00-14:00.

Location: Room G74, Philip Lyle Building.

Map: http://www.info.rdg.ac.uk/maps/maps-display.asp

Abstract:

Over the last five decades electronics has moved from analog to digital form. In this digital era satellite TV, internet and video conferencing are basic ‘needs’. Digital videos are more easily manipulated, stored and transmitted over the network, than analog videos. But, to transmit good quality digital video on limited bandwidth communication channels, video data needs to be compressed. Video compression (coding) tries to exploit the considerable amount of data redundancy and irrelevancy that exists in digital video. Motion compensation is the key to good compression however, computing motion compensation vectors consumes up to 80% of computational power of the encoder.

The aim of the work presented in this seminar is to improve the speed-up of the motion compensation process without introducing any distortion or loss of quality, while reducing the processing time required for video encoding. An algorithm based on a new hashing technique that achieves good quality video and speed up is introduced. Theoretical analysis of this new algorithm shows a speed-up of 1.1 for existing exhaustive search block matching algorithms (BMA) and a speed-up of 1.8 for logarithmic search BMA. Reuse of the hash table provides significant increases in speed-up, with 22.5-24.5 for exhaustive search BMA and 3.5-3.8 for logarithmic search BMA.