Robin Atkins

Color signal encoding for high dynamic range and wide color gamut based on human perception

A new EOTF based on human perception, called PQ (Perceptual Quantizer), was proposed in a previous work (SMPTE Mot. Imag. J 2013, 122:52-59) and its performance was evaluated for a wide range of luminance levels and encoding bitdepth values. This paper is an extension of that previous work to include the color aspects of the PQ signal encoding. The efficiency of the PQ encoding and bit-depth requirements were evaluated and compared for standard color gamuts of Rec 709 (SRGB), and the wide color gamuts of Rec 2020, P3, and ACES for a variety of signal representations as RGB, YCbCr, and XYZ. In a selected color space for any potential local gray level 26 color samples were simulated by deviating one quantization step from the original color in each signal dimension. The quantization step sizes were simulated based on the PQ and gamma curves for different bit-depth values and luminance ranges for each of the color gamut spaces and signal representations. Color differences between the gray field and the simulated color samples were computed using CIE DE2000 color difference equation. The maximum color difference values (quantization error) were used as a metric to evaluate the performance of the corresponding EOTF curve. Extended color gamuts were found to require more bits to maintain low quantization error. Extended dynamic range required fewer additional bits in to maintain quantization error. Regarding the visual detection thresholds, the minimum bit-depth required by the PQ and gamma encodings are evaluated and compared through visual experiments.

Influence of Screen Size and Field of View on Perceived Brightness

We present a study into the perception of display brightness as related to the physical size and distance of the screen from the observer. Brightness perception is a complex topic, which is influenced by a number of lower- and higher-order factors—with empirical evidence from the cinema industry suggesting that display size may play a significant role. To test this hypothesis, we conducted a series of user studies exploring brightness perception for a range of displays and distances from the observer that span representative use scenarios. Our results suggest that retinal size is not sufficient to explain the range of discovered brightness variations, but is sufficient in combination with physical distance from the observer. The resulting model can be used as a step toward perceptually correcting image brightness perception based on target display parameters. This can be leveraged for energy management and the preservation of artistic intent. A pilot study suggests that adaptation luminance is an additional factor for the magnitude of the effect.

Realtime Cross-Mapping of High Dynamic Range Images

The ability to deliver high dynamic range (HDR) and wide colour gamut (WCG) imagery is crucial to next generation broadcast. It is a key feature of both DVB UHD-1 Phase 2 and the latest ITU-R recommendation: BT.2100. While this is an important step towards the creation of broadcast HDR-WCG systems, if HDR-WCG production is to be deployed commercially, it is necessary to use a mix of both conventional standard dynamic range (SDR) and HDR cameras in a single HDR-WCG production. It is also necessary to derive a high quality conventional ITU-R BT.709 (SDR with gamma nonlinearity) programme for regular contribution and transmission. Additionally, it is necessary to cross-map SDR programmes, interstitials and adverts into an HDR-WCG service for transmission. This paper describes the techniques that have been developed to perform these transforms to meet broadcast production standards in real-time. These techniques are built on the experience gained in the creation of the first fifty HDR theatrical releases, as well as trials with HDR broadcast productions. Finally, the operational practices to ensure consistency in HDR-WCG production, high quality programme interchange, and a pleasing viewer experience are examined.

Prediction of HDR quality by combining perceptually transformed display measurements with machine learning

We present an approach to predict overall HDR display quality as a function of key HDR display parameters. We first performed subjective experiments on a high quality HDR display that explored five key HDR display parameters: maximum luminance, minimum luminance, color gamut, bit-depth and local contrast. Subjects rated overall quality for different combinations of these display parameters. We explored two models | a physical model solely based on physically measured display characteristics and a perceptual model that transforms physical parameters using human vision system models. For the perceptual model, we use a family of metrics based on a recently published color volume model (ICT-CP), which consists of the PQ luminance non-linearity (ST2084) and LMS-based opponent color, as well as an estimate of the display point spread function. To predict overall visual quality, we apply linear regression and machine learning techniques such as Multilayer Perceptron, RBF and SVM networks. We use RMSE and Pearson/Spearman correlation coefficients to quantify performance. We found that the perceptual model is better at predicting subjective quality than the physical model and that SVM is better at prediction than linear regression. The significance and contribution of each display parameter was investigated. In addition, we found that combined parameters such as contrast do not improve prediction. Traditional perceptual models were also evaluated and we found that models based on the PQ non-linearity performed better.

Objectively Evaluating High Dynamic Range and Wide Color Gamut Color Accuracy

Color accuracy plays a vital role in the motion picture and television industry, where it is analyzed all the way from lighting during production, through compression, to television sets in the home. The most common metric used for evaluating color quality is ΔE2000. This metric was designed using standard-dynamic-range signals and is based upon the perceptually nonuniform L*a*b* color space. It has not yet been well tested or optimized for high-dynamic-range and wide-color-gamut signals. In this paper, we will present an extension to existing color difference data sets that includes data from high-dynamicrange and wide-color-gamut displays. From these results, we propose an alternative color differencing metric, ΔIC T C P , which improves the performance and efficiency of objectively evaluating color accuracy.