75‐2: Invited Paper: Large Scale Subjective Evaluation of Display Stream Compression

Abstract

VESA Display Stream Compression (DSC) is a light-weight codec designed for visually lossless compression over display links. Such high-performance algorithms must be evaluated subjectively to assess whether the codec meets visually lossless criteria. Here we present the first large-scale evaluation of DSC1.2 according to ISO/IEC 29170-2.

85‐1: Visually Lossless Compression of High Dynamic Range Images: A Large‐Scale Evaluation

“…As described above, the HDR content was selected to challenge the codecs, in spite of this both DSC 1.2a and VDC-M performed very well. This finding is consistent with previous series of experiments using SDR images [5].…”

“…Thus, in addition to objective tests, subjective trials are needed to determine if the output of a particular codec is visually lossless (observers cannot differentiate between the compressed and original content) [3]. To date, DSC1.2a has undergone extensive subjective testing [4,5] using 8-bit per color channel imagery, but not with HDR content and displays. The increased luminance dynamic range (to 1,000 nits) challenges display stream compression, particularly in high-resolution formats, and may enhance the visibility of artefacts.…”

“…VESA DSC 1.2a and a pre-released version of VDC-M compression were applied to the full frame image and a 400 x 600 pixel region was cropped from both the original (reference) and compressed version. The location of the crop was selected to constrain the observers' attention within a region of interest [5]. Each pair of images was displayed at the centre of the screen, with a background luminance of 0.48 cd/m2.…”

“…criteria for visually lossless performance is that no observer could correctly identify the reference image on more than 75% of the trials. As outlined previously [5] this criteria is extremely sensitive to outliers. A modified version of the criteria adopted by VESA focuses instead on the mean and standard deviation of the responses (see Figure 4 legend).…”

P‐31: A Statistical Paradigm for Assessment of Subjective Image Quality Results

“…The ISO/IEC 29170-2 document provides a reporting guideline with recommended descriptive statistics and graphs for analyzing and presenting results. This method was employed recently to conduct a large-scale evaluation of the VESA Display Stream Compression 1.2 [4]. While such an approach is useful for assessing the visibility of artifacts over a population of users.…”