Research Of Realistic Display Technology In Mixed Reality System

Mixed Reality technology becomes more conspicuous with the rapid development of technical renovation,and display technology is one of the most important parts of it.Mixed reality systems require highly virtual-real fusion.However there are usually multiple display devices in one system.Different application environments and display devices lead to the color inconsistency among devices.The color inconsistency problem will sharply reduce the sense of reality and immersion,so we need a high-quality color calibration algorithm to achieve color consistency.At the meantime,the algorithm should be fast with simple calibration steps to enhance the practicability.Traditional color calibration algorithms are generally based on 3D look-up table or linear model.In practice,the 3D look-up table require establishing lots of color corresponding relationships,which means it will occupy a large amount of memory,and the query efficiency will be low.Consequently,the 3D look-up table method is impractical to enforce.Then linear model uses linear function to describe coloring mechanism,establish linear color relationship,but this kind of method is based on the assumption that the spectral function of display device has been linearized.In practice,spectral functions of certain display device are nonlinear,which reduces the calibration quality of linear model.In addition,both 3D look-up table and linear model needs a lot of color samples,results in complex calibration steps and long calculation time.And their practicability is undermined as well.The research of this thesis is focus on the color inconsistency problems.After profoundly analyzing the inconsistency reasons,we propose a nonlinear calibration algorithm based on thin-plate spline with higher quality and faster calibration.Meanwhile an improved local adaptive color preprocessing algorithm is presented to solve the color anomaly problem when calculation breaks the bounds,which preserves more color details and have real-time operating efficiency.The main works of this thesis are as follow:(1)We made an in-depth optical analysis on color inconsistency among display devices,and existing calibration algorithms are compared.We found the reasons come from internal and external disturbances.The internal disturbances are the differences on spectral functions among different devices,and the external disturbances include the ambient light and display surface.We also compared different existing calibration algorithms,and inferred the nonlinear model is most practical.(2)We proposed and realized a nonlinear color calibration algorithm based on the thin-plate spline.We firstly compared the common spatial interpolation methods,then derived that the thin-plate spline is most suitable on color interpolation as it can fit a smooth surface with fully respect on original data.Then we expanded it to three dimensional and proposed a fast and high quality calibration algorithm with GPU parallel acceleration.(3)We proposed an improved local adaptive color preprocessing algorithm to solve the color anomaly problem when calculation breaks the bounds.We improved a previous algorithm,proposed a local chrominance and brightness preprocessing algorithm.We used local linear scaling to adjust the colors into suitable range,which reduces the color variation,keeps more color details.For algorithm efficiency,we put our algorithm in GPU and use the coarse sampling to make our algorithm faster,which attains the real-time preprocessing.(4)We did some experiments to verify our algorithms.For local color preprocessing algorithm,we did some statistical analysis on color variation and algorithm speed.It is indicated that our local color preprocessing algorithm has less color variation than the original algorithm,preserves more color details,and is faster than the original algorithm achieving real-time requirement.For color calibration algorithm,we designed three progressive conditions: different devices,different ambient light and different display surface.Experimental results show that our color calibration algorithm has higher calibration quality and faster speed than former ones.