색수차 왜곡 현상의 디지털 프로세싱에 관한 연구

Abstract

This thesis deals with digital analysis and processing of chromatic aberration, which appears in every lenses used optical devices.To present a comprehensive study of chromatic aberration, we analyzed the reflectance property of optics and examine the merits and the restrictions of conventional reduction algorithms by classifying the approaches into two large categories: the mechanical approach and the image post-processing approach.We first present a new quantitative measure for chromatic aberration and explain a detecting process of color fringes. In spite of many effective algorithms about minimizing chromatic aberration, efficient quantitative measure is not defined yet so that the performance comparison of each algorithm is un-realizable. To calculate the defined quantitative measure, color fringes, which are created due to chromatic aberration, are detected directly from a single degraded image. By analyzing the color behavior on edges that do not show chromatic aberration, a range limitation property for color difference signals are suggested. When pixels violate this proposed property, they can be considered as color fringes. So by checking the verification of color difference property for edges, degraded pixels due to chromatic aberration are detected directly from a single degraded image. The detection process is composed of two steps, first is a detection process for edge transition region and next is a simple difference comparison operator. Then the measure is defined in a ratio of detected color fringe pixel number to whole image size. Because the color fringe detection process does not need any a priori knowledge, the proposed measure can be used in any conventional corrected results.In addition, we introduce and present a new image post-processing method that removes color fringes. Conventional image post-processing methods all tried to model and estimate the aberration kernels first and then remove the color fringes by operating inverse kernels to the whole image area. Because the inversing process is applied to whole image area, new area with no degradation effects could be paradoxically degraded after the correction process. However, the proposed correction method is only implied to the color fringe detected regions so that the method prevents an occurrence of new degradation which can be created in un-degraded area. Additionally, the computation complexity is also reduced. Because the color differences in the degraded pixels, which violate the color difference property for edges, are mainly responsible for color fringes, the correction method is operated to adjust the color difference values to be all remain inside the proper range condition. Experimental results demonstrate the performance of the proposed method is effective.