高速被動式自動對焦技術

High-Speed Passive Autofocus Technique

黃嘉慶(Chia-Ching Huang)

指導教授 ： 陳傳生

元智大學/工程學院/機械工程學系/碩士(2002年)

摘要

常見於數位相機的主動式自動對焦技術，必需藉助測距裝置，測量物件相對於鏡頭的位置，據以調整鏡距以達成對焦目的。被動式自動對焦技術不需要測距裝置，利用鏡頭擷取的數位化影像，計算所謂的清晰度尺度(sharpness measure)後，改變鏡頭焦距或是物距，尋找清晰度尺度達到最大值的成像位置以呈現最清晰的畫面。
迄今已有許多清晰度運算法相繼被提出，但尚未有完善的清晰度運算法能適用於各種環境，例如低照明度環境、高反光度或多段差物體、抑或是計算速度慢。為改進此缺點，必須發展出運算速度快，且適用性更廣的清晰度運算法；此外，對焦點的搜尋法，也是被動式自動對焦的重要發展課題。提高對焦速度可以提高如三次元測量的工作效率。現存的對焦點搜尋法，速度過慢，無法快速尋找出最佳對焦位置。
本論文提出兩種清晰度運算法，利用頻域轉換的方式，推導出離散餘弦轉換DCT(Discrete Cosine Transform)方法；使用空間遮罩，以差分的YZU方式。DCT與YZU的方法和現存的清晰度運算法相比，適用性更廣，影像對焦具有較高的解析度，對於多段差物件有更佳的分辨能力。而YZU的方法能兼具運算速度快的優點，實用性更高。
現存的二分及Fibonacci 搜尋法，必須逐次縮小搜尋範圍，驅動鏡頭在對焦點附近多次來回搜尋，逐次逼近最佳對焦點。本文提出的高速動態對焦搜尋法，不但節省搜尋時間，而且可適用在低照明度及多段差物體。
藉由結合YZU與高速動態對焦搜尋法，設計一些搜尋對焦點方法，實驗結果顯示，此法不但對焦快速，而且對焦準確度高，亦可應用於不同平面的對焦。證明本論文提出的DCT與YZU結合高速動態對焦搜尋法，可達成高速被動式自動對焦技術。

關鍵字

被動式自動對焦技術 ； 高速動態對焦搜尋法 ； 自動對焦 ； 清晰度尺度 ； 離散餘弦轉換 ； YZU ； 對焦點搜尋法

並列摘要

When applying the active autofocus technique, which has been used pervasively in digital cameras, the rangefinder in this case plays an important role in detecting the distance of the object from the camera and in helping the camera to stay in focus. Instead of using the rangefinder to help with focusing, the passive autofocus technique relies on the digitized image to compute the called sharpness measure and to help change the focal length or the distance of the object when searching for the point where the sharpness measure is maximized and the sharpest possible image is performed.
By now, quite a few different calculations of sharpness have been proposed, yet none of them could work perfectly under any circumstances. Failing under low intensity of light, being confused with multiple layers, having little resistance toward reflecting light or simply being slow in calculation are all the problems posed in certain methods. Hence, a great time saver with wider range of application should be developed. Focus Searching Method also has an important part on the way of developing passive autofocus technique. Unfortunately, the existing searching methods have been too slow in finding the best focus point. For example, to improve performance of 3D's measurement, we can increase the speed of the searching focus.
Two sharpness calculations have been suggested in this thesis. One is Discrete Cosine Transform (DCT), which is derivative by using frequency transform; the other is YZU, which is derivative by employing the spatial mask and gradient method. Comparing to the existing calculation, both DCT and YZU can reach a wider range when applying, have a better resolution of the image, and have a greater capability in discerning objects with multi-layers. Yet YZU is more practical and outstanding when compacted with higher speed of calculation.
The existing Binary and Fibonacci searching method need to be manipulated by reducing the searching field gradually and driving the lens back and forth around the focal point yet eventually getting closer and closer to it. In this case, High-Speed Dynamic Focus Searching Method is recommended in order to save the time of searching that can use under low intensity of light and multiple layers.
A couple of searching methods have been derived from the combination of YZU and High-Speed Dynamic Focus Searching Method. Actually, the experiment has shown the combination to be faster and more accurate in focusing that can also apply in the object with multi-layers. It proved true that the combination of DCT, YZU and High-Speed Dynamic Focus Searching Method proposed in this thesis can complete High-Speed Passive Autofocus Technique.

並列關鍵字

Passive Autofocus Technique ； High-Speed Dynamic Focus Searching Method ； Autofocus ； Sharpness Measure ； Discrete Cosine Transform (DCT) ； YZU ； Focus Search Method

參考文獻

[1]. Subbarao, M., and Tyan, J. -K., “The Optimal Focus Measure for Passive Autofocusing and Depth-from-Focus,” Proceedings of the SPIE Conference on Videometrics IV, Vol.2598, pp.89-99, Philadelphia, Oct. 1995.

[2]. Adams, A., “The Camera,” New York Graphic Society, Boston, 1980.

[6]. Subbarao, M., and Tyan, J. -K., “Selecting the Optimal Focus Measure for Autofocusing and Depth-from-Focus,” Proceedings of the IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 20, No. 8, pp. 864-870, August 1998.

[8]. Ng Kuang Chern, N., Poo Aun Neow, and Ang, M.H., Jr., “Practical Issues in Pixel-Based Autofocusing for Machine Vision,” Proceedings of the IEEE International Conference on Robotics and Automation, Vol.3, pp.2791-2796, 2001.

[9]. Horn, B. K. P., “Focusing,” Technical Report AIM-160, Massachusetts Institute of Technology, May 1968.