基於局部圖樣之強健式內/外插影像縮放技術

Abstract

有別於一般單純的影像內插放大或者是放大後再另外強化，本論文提出一個可同時對影像進行放大及強化的方法。此方法透過對局部影像的分析與內建140個圖樣的比對，選擇性地對不同區域作不同的內/外插處理。針對各區域的不同性質而交互進行內插補點或是外插補點，可保有原始影像中的低對比區域之平滑度與各區域周界之銳利度。實驗結果顯示，上述放大後影像於平滑度與銳利度的平衡，提供了使用者絕佳的觀賞品質。並且由於演算法運算簡單且暫存記憶之有限，本論文所提出之方法未來將有利於硬體的實作。

Unlike image scaling through simple interpolation, which may be followed by some image enhancements, this thesis proposes a method which will scale and enhance the images simultaneously. The proposed approach achieves image scaling by performing either (i) interpolation or (ii) extrapolation according to the analysis of local image region based on 140 4*4 image patterns. While the former helps to keep the smoothness within low contrast image regions, the sharpness along border of such regions will be maintained by the latter. Experimental results show that the balance between smoothness and sharpness can indeed provide fine image quality of the scaled images thus obtained. Because of simple arithmetic operations and limited memory buffers, the proposed approach will be favorable in hardware implementation in the future.