eduzhai > Applied Sciences > Engineering >

Nonlinear Transform Coding

  • king
  • (0) Download
  • 20210505
  • Save

... pages left unread,continue reading

Document pages: 17 pages

Abstract: We review a class of methods that can be collected under the name nonlineartransform coding (NTC), which over the past few years have become competitivewith the best linear transform codecs for images, and have superseded them interms of rate--distortion performance under established perceptual qualitymetrics such as MS-SSIM. We assess the empirical rate--distortion performanceof NTC with the help of simple example sources, for which the optimalperformance of a vector quantizer is easier to estimate than with natural datasources. To this end, we introduce a novel variant of entropy-constrainedvector quantization. We provide an analysis of various forms of stochasticoptimization techniques for NTC models; review architectures of transformsbased on artificial neural networks, as well as learned entropy models; andprovide a direct comparison of a number of methods to parameterize therate--distortion trade-off of nonlinear transforms, introducing a simplifiedone.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...