傅楸善臺灣大學：資訊工程學研究所莊曜誠Chuang, Yao-ChengYao-ChengChuang2007-11-262018-07-052007-11-262018-07-052006http://ntur.lib.ntu.edu.tw//handle/246246/54005在此篇的論文裡，我們提出了一種高速的方法來解決暗電流的問題。 傳統的方法裡，每當我們照一張相，我們必須照另一張鏡頭蓋起來的暗電流影像，且這兩張的曝光時間是相同的。現今所有的相機公司都使用傳統的方法來解決暗電流的問題。然而，現代人想要他們的數位相機可以在任何場景任何環境拍照任何東西。長時間曝光會使得傳統方法產生缺點。當使用者拍了一張長時間曝光的照片，他們必須額外等待相同的時間來拍照暗電流影像。這是相當沒有效率的等待。此外，這等待也限制了使用者，使得使用者無法連續長時間的拍照。因此，傳統的方法尚有許多改善空間。 我們所提出的高效率暗電流消去法藉由分析頻率空間圖形可以解決上述問題。我們的方法能在瞬間產生適當的暗電流影像，且影像品質幾乎可以達到傳統方法相同的效果。In this thesis, we propose a high speed method to do dark frame subtraction. In traditional method, when we take one picture, we should take another dark picture (dark frame) with lens hood covered, and they take the same exposure time. Currently, all camera companies use this traditional method to solve the problem of dark current signals. However, people want their digital still cameras to capture everything in every situation under every environment. Long-time exposure makes traditional method defective. When users take a long-time exposed picture, they must spend another equal time to wait for this picture’s dark frame. It is inefficient for the waiting. Furthermore, the waiting restricts users from consecutive long-time captures. Therefore, this traditional method should be improved. Our proposed high speed dark frame subtraction can solve this problem by analyzing patterns in frequency domain. Our method can generate suitable dark frames in an instant, and the image quality by our algorithm almost reaches that of the traditional method.1. Introduction 1 1.1 Motivation 1 1.2 The Effect of Dark Current Signals 2 2. Dark Frame Subtraction 5 2.1 Dark Frame’s Structure 5 2.2 Traditional Method 7 2.3 My Improvement 7 3. Used Tools 9 3.1 Regression 9 3.1.1 Linear Regression with Degree One 9 3.1.2 Polynomial Regression 11 3.2 Gaussian Distribution 13 3.2.1 Statistical Model 13 3.2.2 Gaussian Low Pass Filter 15 4. High Speed Method in 15-Second Exposure Time 18 4.1 Outline 18 4.2 Environment 18 4.3 Computed Extreme Value Pixels 24 4.4 Computed Fixed Pattern Area Pixels 32 4.5 Add Noise and Form an Artificial Dark Frame 46 5. High Speed Method in Different Exposure Time 52 5.1 Relations between Exposure Time and Extreme Value Pixels 52 5.2 Relations between Exposure Time and Fixed Pattern Area Pixels 54 5.3 Relations between Exposure Time and Noise 58 6. Experiment Result 61 6.1 PSNR Comparison 61 6.2 SNR Comparison 62 6.3 Processing Time and Needed Space Comparison 63 6.4 Human Eye’s Feel 65 6.5 Conclusion and Future Work 69 Reference 711764667 bytesapplication/pdfen-US電荷藕合裝置暗電流消去CCDsensordark framethesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/54005/1/ntu-95-R93922087-1.pdf