영역의존 저역통과필터에 기초한 적응형 고능률 디인터레이싱 방법

Abstract

The interlaced video format widely used in recent broadcasting systems which is required to be converted into the progressive format in display systems such as high-definition television (HDTV), PC monitors, liquid crystal displays (LCD) and plasma display panels (PDP). This conversion process is called de-interlacing which can be roughly divided into two categories: spatial interpolation methods and temporal interpolation methods. Each technique has its own advantage and limitation. Spatial methods only use the pixel values of the current frame, so they have relatively light computational complexity and are easy to implement. Certainly, they have been widely used in practice, especially to process large video data in real time. However, they may produce some artifacts in edge and moving areas. The temporal de-interlacing methods utilize the information from adjacent fields to fill in missing lines. Although they provide improved performance among various de-interlacing methods, they require high computational complexity. Moreover, in motion compensated de-interlacing methods, inaccurate motion estimation may cause artifacts which degrade visual quality. In this thesis, two efficient spatial de-interlacing algorithms are presented to improve the conventional methods. Both of them are based on the idea of region segmentation, in order to suit the remedy to the case. The first proposed algorithm considering the influence of bilateral filter. In this method, the image is firstly decomposed into base layer and detail layer by bilateral filter. Then different interpolation methods are applied on different layers since their disparate characteristics. At last, the interpolated image is obtained by adding the two layers. Experimental results show that the subjective quality of the interpolated images is substantially improved by using the proposed algorithm compared with that of using conventional interpolation algorithms. The second proposed algorithm introduces a high performance de-interlacing algorithm based on diverse interpolation filters for the corresponding regions. Specifically, due to most of an image with less complex directional correlation, so the 6-tap fixed Wiener interpolation filter only use for the most complex region, but for the smooth and regular edge region, much more correlated filter such as 2-tap or 4-tap fixed Wiener interpolation filters should be employed instead. According to the experimental results, the proposed algorithm has achieved noticeably good performance with low cost.