Tone mapping-based high-dynamic-range image compression: study of optimization criterion and perceptual quality

Abstract

International audienceWe study the issue of quality assessment in tone mapping- based high-dynamic-range (HDR) image compression. In this, there are two stages at which a decision should be made regarding perceptual vis- ual quality: (a) for finding the optimal parameters of the dynamic range reduction function so that the visual quality is maximized, and (b) visual quality judgment of the decompressed image. We first investigate two objective optimization criteria, namely mean squared error and structural similarity index measure, toward optimization of a tone mapping model- based HDR image compression method. We then conduct a comprehen- sive subjective study to evaluate the visual quality of the compressed HDR images. Therefore, we consider both objective and subjective aspects for HDR image compression. To our knowledge, no systematic and compre- hensive studies exist in the current literature which shed light on the issue of quality assessment in HDR compression. So this study brings in new knowledge and perspective for the relatively less investigated topic of HDR compression from the view point of perceptual quality. We further evaluate the prediction performances of four objective methods on the 140 compressed HDR images that have been subjectively rated

An extensive performance evaluation of full-reference HDR image quality metrics

“…The capability of both kinds of fidelity metrics to predict viewers' mean opinion scores (MOS) has been assessed in a number of recent subjective studies using compressed HDR pictures (Valenzise et al 2014;Hanhart et al 2015a;Narwaria et al 2013Narwaria et al , 2012a. Nevertheless, the results of these studies show sometimes discrepancies in their conclusions about the ability of these metrics to yield consistent and accurate predictions of MOSs.…”

High Dynamic Range Video Compression Exploiting Luminance Masking

“…The results of [14] for example suggest that, at higher bitrates, HDR-VDP-2 scores do not increase in proportion to subjective scores. Moreover, as stated above, HDR-VDP-2 does not consider color information and therefore artifacts (e.g.…”

On improving the pooling in HDR-VDP-2 towards better HDR perceptual quality assessment

“…150 conditions = 10 reference images × 15 conditions per reference image, to be evaluated by subjects. The keen reader is also referred to our previous work 18 for further details.…”

“…As mentioned, currently there is no publicly available HDR database with subjective quality ratings. Thus, the performance of HDR-VDP-2 and even conventional LDR metrics has not been evaluated with HDR content except our previous study 18 in which we evaluated the performance for JPEG compressed HDR images. In this paper, we further validate the performance of objective methods on HDR images affected by JPEG 2000 compression errors as well as distortions due to tone mapping.…”

“…The mean opinion score (MOS) for each stimuli was obtained by averaging the scores for that stimuli from all the observers. The keen reader is also referred to our previous works 18,19 for further details on the test material preparation and the subjective experiments.…”