本論文探討如何改進顯示器的顯色精準度，尋求一有效的校正模 式，期能大幅改進顯示器的顯示效能。首先，針對現行顯示器的顯色 效果，進行實際量測、並分析其光度與色度的變化情形，從而建立其 量化模型，藉由操控用灰階值與顯示器效能表現之間的光度、色度精 確關係模型，提供顯示器光度與色度精確校正之應用。現行的校正方 式，僅考慮光色分離的方式，利用三原色混色的原理做固定三原色的 不同光度混色，最後再輔以線性轉換，試圖用以解決混色預測效果與 實際呈色的差異，最後仍存在明顯不等的色差，使得每部顯示器無法 完全達到相同的顯色表現。本研究提出線性疊加混色模型與光譜主成 份混色模型，除了有效建立精準的光色混色模型，同時也藉由數據分 析與模型建立的過程，推測出現行三原色混色模型疏失之處，從而提 出顯示器架構所引進的三原色互相漏光透色之機制，造成當前顯示校 正程序的高複雜度，也因此找到合理的精準預測模型，大幅而有效地 降低各色階之色差，達成顯示器顯色之精確校正的目的。於此，相對 於線性疊加混色模型，光譜主成份混色模型可提供較高的光色精準度， 但其程序較為複雜。;The purpose of this thesis is to explore for an effective calibration process to highly improve the performance of displays. In such a way, the calibrated displays can present high quality contents with high accuracy both in photometric and in chromatic evaluation. Firstly, we have executed photometric and chromatic measurements on the performance of displays. After analyzing the measured data, several photometric and chromatic models between the controlling gray levels and display outputs of luminance and chromaticity have been established for each one of the primary color channels. With these photometric and chromatic models, the high accuracy calibration on a display is approached both in the photometric and the chromatic performance. Currently, the calibration process for a display is based on the color mixing among three fixed color primaries. By a linear transformation, the color difference between the color mixing prediction and the realistic performance of a display is expected to be diminished. However, we found a display does not work in such a simple way. That is why every display has its own exclusive performance even for the same contents. Here, we propose two effective models for the color mixing in displays. One is the linear superposition mixing model, and the other is the spectral principal component mixing model. Both of these two ways can effectively solve the unneglectable color shiting problems in displays. Accordingly, the leaking primary color can be well evaluated such as to have accurate calibration. The spectral principal component mixing model provides with much better accuracy than the linear superposition mixing model, while with more complicated procedures.