A subjective image quality assessment dataset of color graded inverse tone-mapped HDR images

A subjective image quality assessment dataset that includes quality scores of HDR images generated by nine inverse tone mapping methods. The images in the dataset show a wide variety of artifacts commonly present in dynamic range expanded HDR images. Twelve image pairs comprised of an LDR image and its corresponding HDR version were used to conduct the subjective assessment study. These images contain scenes with a wide range of light conditions,  representing challenging situations for dynamic range expansion methods.

The image quality dataset includes subjective quality scores for 108 inverse HDR images obtained by the different dynamic range expansion methods, scaled in Just Objectionable Differences (JODs). In addition, it includes the raw data from pairwise comparisons obtained from subjective experimentation. The raw data is composed of 6480 trials collected from 15 human observers.

Files included

List of images used in our experiments (images.csv).

LDR images used as input (ldr.zip).

HDR images used as reference (hdr.zip).

Inverse tone-mapped HDR images evaluated in our study (hdr_itmo.zip).

Pairwise comparison results and JOD scores (subjective-scores.zip)

The objective quality scores of the inverse tone-mapped HDR images, computed by each quality metric assessed (objective-scores.zip).