The image quality of flat panel displays and image sensors has become increasingly important due to recent diversification of their uses and their rapid diffusion. In defect inspections of those image devices, particularly in the Mura (luminance nonuniformity) inspection, human sensory inspection has been mainly introduced into the manufacturing process. Recently, the introduction of automatic defect inspection has been started in earnest. However, the varieties of Mura and background conditions make the tuning process for achieving the demanded accuracy difficult. In this paper, we report our subjective evaluation experiments to obtain relations between JND (just-noticeable-difference) contrasts and various conditions of Mura, such as area, background luminance, luminance distribution, and polarity. On the basis of those experiment results, we propose a model based on the MTF (Modulation Transfer Function) of the human visual system, which can calculate the JND contrast under various conditions. In addition, we describe the validity of a quantitative method, in which we quantify Mura levels by using the proposed model.
View full abstract