自動偵測振鈴現象於復原模糊人臉影像

Abstract

利用盲卷積進行復原模糊影像的已經廣泛研究了很長時間，但是一般的解決方法面對任何模糊影像仍然是一個巨大的挑戰。在這篇論文中，我們提出了一種新的人臉模糊影像復原方法，根據溫納濾波和辨識技術。嘗試數種不同的模糊圈半徑用於溫納濾波器和區域對比增強，更進一步加強被復原影像的紋理。我們所提出的系統可以自動確定較好的卷積模糊圈半徑，即是當我們在復原模糊影像時可以避免產生振鈴現象。

Restoring blurred images by blind deconvolution has been extensively studied for a long time. But a general solution for deblurring any out-of-focus images is still a big challenge. In this paper, we present a new face image restoration approach based on Wiener filter and pattern recognition techniques. By trying several radii of circle of confusion (COC) used in Wiener filter and applying local contrast enhancement to further enhance the textures of deblurred images, the proposed system can automatically determine the best deconvolution radius of COC such that the deblurred image has less ringing artifacts.