自動對焦的一個新取樣方法於手機相機

Abstract

我們在這篇論文提出峰值偵測自動對焦用一個新的有效及有效率的取樣表給手機相機。許多的自動對焦演算法基於清晰度值的曲線，但是他們並沒有考慮到取樣位置與影像的特性。我們的取樣表是由影像校正而來的，它符合光學鏡片、感測器響應、雜訊、以及清晰度函數的加成結果。在一個三百萬畫素的手機相機中，此表可將搜尋步數由 404 步減少到 16 步，影像仍然保持清晰。在搜尋中的峰值偵測能夠避免雜訊的干擾即使影像是在低對比的狀況下。更進一步，它能輕易地應用於影像晶片或是手機相機上。

We present an auto focus method with peak detection by a new, effective, and efficient sampling table for phone cameras. Many auto focus algorithms base on focus value curves, and they do not consider the correlation between sampling and image characteristic. Our table is calibrated by images to meet results of optical lens, sensor response, noise, and focus value function. In a 3 mega-pixels camera, the table reduces total steps from 404 to 16, and captured images are still sharp. Proposed peak detection in a search can avoid noisy interference even though scenes are lower contrast. For further progress, they can be applied to image processing chips or mobile phones easily.