2019-08-012024-05-18https://scholars.lib.ntu.edu.tw/handle/123456789/703685摘要：根據影像銳利度的傳統自動對焦技術，雖然能掌握最後準度，大多難以穩定地決定第一步鏡頭應該移動的方向，影響對焦速度。最近常用的一種作法試圖在相機感測器的某些像素位置上配置相位偵測器，以相位差計算物體深度。這樣一來，對焦起步緩慢的問題能夠因為獲得深度資訊而解決。越來越多智慧型手機的相機都採用這種硬體結構，蘋果公司自iPhone 6以來的手機系列就是一個例子。然而，這個技術仍有許多有待解決的問題。其中重要的一項是相位差的估測品質不佳，常受到內在或外在因素的影響。舉例來說，光線昏暗環境下，低光通效率會影響相位感測器的運作，這些情形會造成相機無法仰賴相位差資訊進行對焦。根據上述現象，我們可以知道影像銳利度和相位差各有其優劣勢。因此，我們的研究動機在於分析如何截長補短，把這兩種方法作最佳搭配，產生利基，提昇對焦效能。 本計畫分三年執行，主要內容包括: 1)分析相位偵測技術並評估相位差資訊之可靠度，2)研究如何結合相位差與影像銳利度產生一個更有效的自動對焦方法，3)探討如何在對焦過程中即時克服環境光不完美或鏡頭髒汙等干擾因素對自動對焦造成的影響。<br> Abstract: Conventional contrast detection autofocus (CDAF) methods, although accurate, are often unable to reliably decide the direction of initial lens movement, resulting in a slow start. One of the solutions is to equip cameras with phase detection sensors at a small number of pixels across the image sensor to acquire depth information. The slow start problem can then be solved in a cost effective manner with the availability of such sparse depth information. More and more smartphone cameras are equipped with phase detection sensors. For example, the Apple iPhone series has been featured with phase detection autofocus (PDAF) since the debut of iPhone 6. However, there are still many problems to be solved. One of them is the poor quality of phase difference estimate due to, for example, the imperfection of image quality. Besides, it is hard for phase-detection sensor to operate properly in low light condition due to decreased quantum efficiency. Hence, we cannot solely count on phase difference estimate for autofocus. Given the complementary pros and cons of CDAF and PDAF, it becomes clear that a feasible solution is to optimally combine both phase difference and contrast detection for autofocus. That is the ultimate goal of the proposed research, which in our view has both academic and practical values. The execution of this research project is divided into three years with three key parts: 1) Analyze the characteristics of phase detection and enhance the reliability of phase difference for autofocus, 2) Develop an effective search algorithm that integrates both PDAF and CDAF, and 3) Address issues related to imperfect light condition and the presence of lens dirt to enhance the performance of autofocus.自動對焦相位差Autofocusphase difference