| Currently, the smart mobile devices have been extensively utilized in our daily life and work along with the fast-growing display technologies. Therefore, high IQ (image quality) on these mobile displays has become extremely desirable, in the wake of the raise on the demand of evaluating the display performance for different kinds of mobile devices. However, there are no such evaluation methods for smart mobile devices yet, while the simple comparison on their physical parameters such as color gamut, maximum luminance, and PPI (pixel per inch) cannot provide reasonable quantized scores in accord with the perceptual features of human visual system under multiple viewing conditions for mobile devices. Hence, this study focuses on investigating the display performance and modeling the perceptual IQ for smart mobile displays with different technologies under different lighting environments using various application types of test images. Firstly, the physical and colorimetric characteristics were measured and compared for four mobile displays, and then accurate prediction of tristimulus values XYZ and color appearance values was achieved by colorimetric characterization modeling. Afterwards, large-scale visual data were collected via the well-designed psychophysical experiments on the four test displays under three levels of lighting conditions. Based on the analysis of two types of mobile displays with different physical sizes, the procedure and algorithms for predicting perceptual overall IQ were proposed, and the excellent performance of the overall IQ model was verified by the visual data.To evaluate the display performance on the physical aspect, the important colorimetric characteristics were measured for the four mobile displays with different technologies of TFT (thin film transistor), IPS (in-plane switching), and AMOLED (active matrix organic light emitting diode), by which their forward and inverse colorimetric characterization modeling was deeply investigated. Their physical and colorimetric characteristics were analyzed, including the spectral power distributions of three primaries, TRC (tone reproduction curve), repeatability, color gamut, spacial uniformity, channel independence, and chromaticity constancy. Based on the evaluation of channel independence and chromaticity constancy, the characterization performances of GOG (gain-offset-gamma), S-Curve Ⅰ, S-Curve Ⅱ, PLCC, PLVC, and Masking models for predicting tristimulus values XYZ were evaluated and discussed. Considering the aspects of accuracy, efficiency, and the requirements of the visual experiments, the individually suitable colorimetric characterization model was determined for each of the test displays, which set a solid foundation for the processes of image manipulations and the description of color appearance values. Moreover, an effective colorimetric characterization model, namely, PC (polynomial compensation) model, was proposed for AMOLED displays to cope with their poor performance on channel independence, by taking the channel interaction and non-constancy of the primaries into consideration. The results demonstrated that this PC model was accurate and efficient for practical applications, and it could improve the forward characterization accuracy by a large margin. In addition, a3D-LUT (three-dimension look up table) method with8×8×8sampling was employed and further optimized for the inverse characterization of AMOLED mobile displays.The psychophysical experiments were elaborately designed to meet the several requirements for the final IQ modeling, including the completeness of its procedure, as well as its applicability to various smart mobile devices, lighting conditions, and different types of images. The experiments were carried out on two smart phones (one IPS panel, one AMOLED panel) and two tablet computers (one IPS panel, one TFT panel), in order to involve the different effects of display technology and display size on perceptual IQ attributes. The visual assessments were conducted under the dark surround and two different ambient lighting conditions with the illuminance of500lx and10000lx to simulate the indoor office lighting and outdoor illumination condition respectively, resulting in eight experimental phases in total. Considering the specificity of mobile displays in practice, many rather different styles of test images were adopted in the visual evaluation, including both natural scenes containing familiar memory colors, and other application types of scene images without memory colors, as well as artificial drawings, games, maps, and internet. To provide a wide but realistic range of image variations, all the original test images were rendered using various manipulation methods, involving changes in lightness, chroma, hue, and spatial frequency, respectively. Then, several perceptual IQ attributes of naturalness, colorfulness, brightness, contrast, sharpness, clearness, preference, and overall IQ were estimated by a panel of10observers via the psychophysical method of categorical judgment. It is worth noting that only the test images of natural scenes can be evaluated for the attribute of naturalness. Therefrom, massive visual judgments were obtained, which established an important basis for the comprehensive IQ analysis and modelling in the further work.For each IQ attribute, the raw data of category names assigned by all the observers via categorical judgment were converted to the equal-interval scale values by adopting Case V of Thurstone’s law on comparative judgment. Then, the interrelationships among various IQ attributes were analysed in order to determine the critical factors impacting the overall IQ of smart mobile displays. The Pearson correlation coefficients were calculated for each pair of IQ attributes with all their possible combinations to find their important interactions. As the evaluated IQ attributes would not exist in an isolated manner, factor analysis was performed to deeply seek the potential interactions among the IQ attributes and to find the important attributes significantly affecting the overall IQ. Moreover, the influences of the image contents and the manipulation methods were analyzed, and also the impacts of illumination levels on IQ performances were discussed by the implement of ANOVA (analysis of variance). Herewith, the general scheme of the IQ modeling was established upon the comparison and summation of all the experimental phases for smart mobile devices.Based on the detailed analysis of the visual data, the final model of overall IQ was built in two stages, which could provide satisfactory prediction results in accord with the perceptual features of human visual system. In the first stage, the overall IQ could be expressed separately for natural scenes and other application types of images through the method of multiple linear regression, using a linear combination of its constituent IQ attributes of naturalness, colorfulness, and clearness with their corresponding weighting coefficients. The next stage was to establish the mathematical links from image contents and display parameters towards the key constituent attributes of naturalness, colorfulness, and clearness. The color appearance model CIECAM02was adopted in the description of color appearance values for each pixel, since it could take the influence of lighting factors into consideration. Then, by adding the functions of physical parameters into the calculation of color appearance values, the IQ attributes of naturalness, colorfulness, and clearness could be represented accurately. Accordingly, the overall IQ of smart mobile displays could be predicted through a universal workflow by the adoption of color appearance values, illuminance levels, and several physical parameters such as color gamut and PPI. The proposed model could output the overall IQ for smart mobile displays with different technologies under various lighting conditions, of which the practicability and effectiveness were further verified by the visual data.Finally, the main research contents and innovative achievements of this dissertation were summarized, and the perspectives of the future study were also forecasted. |
PDF Full Text Request