本論文提出一個快速影像式自動對焦演算法,主要是應用鏡頭架構特性,尋找出鏡頭景深(Depth of field),再利用鏡頭解像力(Resolution)概念,以調制轉換方法(Modulation Transfer Method,MTF)做為影像清晰度演算法。經由粗略搜尋,算出影像對比度(Contrast)和銳利度(Sharpness)數值,利用焦距搜尋法,尋找出鏡頭的景深範圍,再由細部搜尋景深範圍內的清晰度數值,進而找出鏡頭成像最清楚的對焦點位置。以調制轉換方法做清晰度演算法可以避免以影像差異值和影像梯度值為基礎的清晰度演算法所造成的局部極大值(Local Maximum)現象,以求得全域最大值(Global Maximum),使對焦範圍移動鏡頭景深範圍,不會造成太大的失焦誤差,接著繼續使用調制轉換方法做精密的細部搜尋,求得景深範圍細部的影像對比度數值,尋找出最大值的數值,求得出真正對焦點。經實驗証明本論文提出之自動對焦技術比一般常用的自動對焦法更為準確且快速。
The paper proposes a fast image based auto focus technique. By applying the lens' architecture, depth of field can be estimated by computing the image clarity based on modulation transfer method. Based on coarse to fine strategy, we first compute the image contrast and sharpness and estimate the lens' depth of field by focus searching method. Then, we can find the accurate focus position via finely calculating the image clarity in searching area of depth of filed. Instead of image differential or gradient method, modulation transfer method can avoid the local maximum problem in image clarity computation. By finding the maximum image contrast in the searching area of depth of field, we can find the true focus point based on modulation transfer method. From the experiments, it is proved that the paper proposed a more accurate and fast auto focus method compared with the other traditional ones.