Lossless Video Compression Technique Using Statistical Inference and Entropy Coding

Rochan Avlur Venkat

Rochan Avlur Venkat
AECS Magnolia Maaruti Pre University College, Bangalore, Karnataka
Title: Lossless Video Compression Technique Using Statistical Inference and Entropy Coding
Video capture technology has witnessed an exponential growth in recent years owing to developments in high fidelity video capture techniques. Unfortunately, high fidelity data is often extremely large and an efficient compression method is critical to store, process and transmit such data. Many of the existing compression methods are lossy, which is unacceptable in such environments. The near-lossless compression algorithm designed in this project uses pixel groups, forming multiple sub-images of the differences between two successive frames. Sub-images contain very low context change between successive frames due to high frame rate. The probability of change between continuous successive sample sub-image values is computed using Bayesian Inference technique and filtered using thresholds. Values are later encoded using Entropy coding techniques. This compression method is designed for performance-critical and large data volume applications, which often require very fast data access without prior decompression and for which a modest compression rate is acceptable.