Color adaptation in video coding

Abstract

A video bitstream is received from an encoder, comprising encoded image portions each having a common form representing components of a channel in a color space. Each of a plurality of the encoded image portions comprises a different set of quantized values of the components, including at least a value of a first of the components being quantized from amongst a first scheme of quantized levels. Further, the received bitstream comprises a look-up table mapping the quantized levels of the first scheme to at least partially de-quantized respective levels. A de-quantizer uses the look-up table received in the bitstream to at least partially de-quantize the different quantized values of the first component in a plurality of the image portions, by converting the quantized values of the first component to at least partially de-quantized values corresponding to ones of the at least partially de-quantized levels of the first scheme.

Description

Background technique [0001] Digital cameras tend to capture images with a high color depth much higher than typically required in practice. For example, some cameras capture samples at a depth of 10 or even 12 bits per R, G, and B channel, giving a total depth of 30 to 36 bits in RGB space. [0002] The human eye on the other hand is generally not able to distinguish these many colors. Based on studies of human vision, it is estimated that a typical human being is able to perceive only about 2 million different colors. That corresponds to a total color depth of about 20 bits (6 to 7 bits per channel). [0003] If the captured data is to be encoded for transmission over a network, high color depth information incurs very high bit rates and a high processing burden in encoding. Similarly, high color depths incur significant memory resources if the data is to be encoded for storage. [0004] For this reason, raw image data captured from cameras is often quantized for the purp...

AI Technical Summary

Problems solved by technology

[0003] If the captured data is to be encoded for transmission over a network, high color depth information incurs very high bit rates and a high processing burden in encoding

Images