Perceptual intra-frame coding for HEVC still picture profile based on invisible signal suppression

In this paper, a perceptual intra-frame coding for high-efficiency video coding under the still picture profile is proposed to improve coding efficiency in terms of subjective quality. The proposed method suppresses invisible signals based on the just notable distortion (JND) model by maintaining su...

Full description

Saved in:
Bibliographic Details
Published in:Signal, image and video processing image and video processing, 2020-07, Vol.14 (5), p.1053-1061
Main Authors: Jo, Hyunho, Lee, Seonoh, Ahn, Yongjo, Kim, Yongwoong, Sim, Donggyu
Format: Article
Language:English
Subjects:
Algorithms
Coders
Coefficients
Computer Imaging
Computer Science
Image Processing and Computer Vision
Multimedia Information Systems
Original Paper
Pattern Recognition and Graphics
Quality assessment
Signal,Image and Speech Processing
Video compression
Vision
Citations: Items that this one cites
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In this paper, a perceptual intra-frame coding for high-efficiency video coding under the still picture profile is proposed to improve coding efficiency in terms of subjective quality. The proposed method suppresses invisible signals based on the just notable distortion (JND) model by maintaining subjective quality with a lower bitrate. For quantized transform coefficients of residual signals, the magnitudes of nonzero coefficients iteratively reduce by 10% and the maximum reduction is decided by the non-noticeable quantized coefficients based on the JND measure with a given quantization parameter. The performance of the proposed algorithm was verified on the HM reference encoder. In the all-intra-configuration of the common test condition, the proposed perceptual encoder achieves a bit saving of 13.87%, on average, and up to 39.52% for several test sequences. The mean difference mean opinion score for subjective quality evaluation is observed approximately 0.64–0.15 with 15 subjects.
ISSN:1863-1703
1863-1711
DOI:10.1007/s11760-020-01639-x