Skip to Main content Skip to Navigation
Conference papers

Statistical Analysis of Residue Signal for Backward Compatible HDR Image Coding

Abstract : Backward compatibility to the existing 8-bit displays for HDR image and video compression is an essential requirement for the wide acceptance of HDR technology by the users. While these previous methods mainly focus on the rate-distortion optimization of base layer, the encoding of the residue in the enhancement layer comparatively has received less attention. In this paper, we investigate whether HDR image coding residue exhibits also local anisotropic characteristics or not as in the case of motion compensated LDR video coding residue. The verification of existence of such characteristics, as an essential initial stage in such a research, form a base for the development of directional encoding methods for HDR residue as well. For this purpose, statistical auto-covariance analysis on HDR image coding residue is performed in this paper. Specifically, we first interpret the resulting auto-covariance model parameters for HDR images and then the behavior of these parameters is analyzed with respect to bitrate, spatial activity and dynamic range as the main variables in HDR image coding. The experiments indicate that the HDR residue exhibits local anisotropic characteristics, which enables the use of directional methods for the encoding of enhancement layer in HDR coding.
Document type :
Conference papers
Complete list of metadatas

https://hal-centralesupelec.archives-ouvertes.fr/hal-02395452
Contributor : Delphine Le Piolet <>
Submitted on : Thursday, December 5, 2019 - 2:55:57 PM
Last modification on : Wednesday, September 16, 2020 - 4:50:19 PM

Identifiers

  • HAL Id : hal-02395452, version 1

Citation

Kutan Feyiz, Fatih Kamisli, Emin Zerman, Giuseppe Valenzise, Alper Koz, et al.. Statistical Analysis of Residue Signal for Backward Compatible HDR Image Coding. 25th Signal Processing and Communications Applications Conference (SIU), May 2017, Antalya, Turkey. ⟨hal-02395452⟩

Share

Metrics

Record views

37