인간 시각계 모델을 사용한 Plasma Display Panel의 동영상 왜곡 최소화

Abstract

While very competitive in the market for large-size display applications, plasma display panels (PDPs) have some drawbacks that need be solved to increase their competitive power. Motion picture distortion (MPD) is one of such drawbacks, which is inherently induced by the subfield driving method of PDPs. PDP adopts the subfield driving method to implement the gray levels of images. However, due to the on-off mismatches on subfields, when the human visual system (HVS) tracks the movement of the object on the PDP, the amount of the light accumulated on the human retina varies and this induces MPD. Therefore, if we minimize the on/off mismatches on subfields used for the gray levels of neighboring pixels, we can minimize the MPD that human recognizes. This dissertation presents a new optimization approach that dynamically changes the subfield codewords for given image, which minimizes the MPD that the human visual system recognizes. The proposed method uses the fact that MPD is more perceivable at object surfaces than at object edges. In addition, it minimizes the on-off mismatches on the subfields used for object surfaces by giving them priority for codeword selection. For the implementation, a new MPD evaluation method which considers the MPD generated by all possible gray level pairs is used. Next, the cost function and the search space of the new codeword mapping optimization method are evaluated using the MPD evaluation method. Finally, a HVS model is applied to optimization to minimize MPD that the human visual system recognizes. The performance of the proposed approach is experimentally verified using Structural Similarity (SSIM), which is the noise metric designed to consider the characteristics of the HVS. The MPD simulation results show that the proposed approach, changing codewords dynamically, improves the qualities of test images by 0.012 on the average and 0.052 maximally in SSIM, compared with the method using the fixed codewords. The experiments of the MPD simulation also illustrate that the proposed approach maintains the image quality well for all test images, providing them the SSIM values over 0.9 when compared with original images.