Oleg Zabluda's blog
Saturday, June 10, 2017
 
What is the best lossless video compression technique with the best compression ratio?
What is the best lossless video compression technique with the best compression ratio?
"""
Matt Mahoney, Developed PAQ and ZPAQ. Maintains the Large Text Benchmark and others. On the Hutter Prize committee. Answered Feb 16, 2016

[...] The information content of video is dominated by noise in the low pixel bits. The noise is both useless and incompressible. You either have to remove it or settle for around 2:1 or 3:1 compression ratio depending on the source. This is still huge. For example, 1920 x 1080 x 30 fps x 8 bits x 3 colors = 1492 Mbits/sec before compression and maybe 500 Mbits/sec compressed. Compare this to 3 to 5 Mbits/sec for typical MPEG-4 HD video.

The simplest lossless compression method would be to subtract neighboring pixels and compress the difference using an adaptive order 0 indirect context model. An indirect context model maps a context (the previously coded bits of the current pixel) to a history of the last several bits, and then to a prediction, which is arithmetic coded. You could get better (but slower) compression by mixing the predictions of several models that include neighboring and prior pixels as context. Developing the model would be experimental: try different contexts and measure the compression.
"""
https://www.quora.com/What-is-the-best-lossless-video-compression-technique-with-the-best-compression-ratio
https://www.quora.com/What-is-the-best-lossless-video-compression-technique-with-the-best-compression-ratio

Labels:


| |

Home

Powered by Blogger