A Quantitative Evaluation Method for Luminance and Color Uniformity of a Display Screen Based on Human Perception

Deep convolutional neural network‐based quantitative evaluation of display non‐uniformity

“…Finally, we calculated the effective error of the proposed method. We also calculated the evaluation values of the Nagamine method, 3 the SEMI method, 6 and our previous method 8 for comparison. Figure 5 shows examples of pseudo multiple mura images, the captured luminance images, and the squared error images evaluated by the proposed method.…”

“…Many methods have been used to overcome this difficulty. [1][2][3][4][5] For example, visual sensitivity characteristic models have been applied to a display image [1][2][3][4] or edge detection and differential filtering 3,5 to emphasize or extract mura and then calculate quantities of specific mura characteristics. Quantitative evaluation of mura is also supported by a few standards.…”

“…However, such approaches have difficulties in optimizing the formulae to evaluate the degree of mura, and even when the formula is optimal, the evaluation must be performed under restricted viewing conditions (e.g., display luminance, environmental illuminance, and viewing distance). For example, the method of Nagamine et al 3 defines the mura edge as one of the numerical indexes in the formula. Extraction of the edge depends on the user's discretion, and it is difficult to extract the edge correctly based on the user's viewing conditions.…”

81‐2: Invited Paper: Neural Network Based Quantitative Evaluation of Display Non‐Uniformity Corresponds Well with Human Visual Evaluation

“…To overcome the difficulty, many researches had conducted. For example, applying visual sensitivity characteristic to a display image [1, 2, 3, 4,] or edge detection and differential filtering [3,5] to emphasis or extract Mura, then, calculating feature quantities of Mura have performed. However, these approaches had difficulties in optimizing the feature quantity to evaluate Mura degree or must be performed under restricted conditions even if the feature quantity was decided well.…”

68‐4: Speedy and Quantitative Evaluation of Luminance Non‐Uniformity Based on Deep Neural Networks

“…Accordingly, much research has been conducted into finding a way to uniformly measure Mura. For example, applying visual sensitivity characteristics to a display image [1, 2, 3, 4,] or edge detection and differential filtering [3,5] to emphasize or extract Mura, then calculating Mura feature quantities have been tried. However, these approaches had difficulties.…”