現在大多數人的生活當中，人手一隻可拍照的手持式裝置已經成了生活必需品。從以前傳統的底片相機，到後來的傻瓜相機，再到現在的數位相機，甚至現在的人已經不隨身帶著相機了，取代而之的是可拍照並且可即時將所拍的照片上傳到網路上的智慧型手機。
雖然智慧型手機的高便利性的好處，但因為手機設計往往以輕巧型為主要，就性能而言，確實無法與大尺寸像素的傳統／數位相機比較。　在遇到高反差場景(同時存在高亮度細節與低亮度細節影像)，比如說太陽直射，但同時又有大樹下的陰影，反而無法呈現出如人眼所看到的清晰影像。為了解決此的問題，需要使用高動態範圍影像(High dynamic range, HDR)的合成技術。
目前已經存在有非常多種類HDR的合成技術。 比較常見的分為兩類，前陣子較流行的曝光融合技術(Exposure fusion)與傳統的色調映射(Tone mapping)技術。　前者以高質影像表現，較不容易出現一些不良現象，但耗費較高的記憶體和運算單位資源且運算複雜度高；後者為運算複雜度低，結果比較Exposure fusion方法來說，在特定場景下容易出現如光暈輪(Halo)等現象或細節被壓縮的情況。
本論文研究主要是以Tone mapping技術為主要研究方向，希望能有較高的效能，在低成本的影像傳感器(CMOS image sensor)上實現即時的影像顯示，並且配合既有的影像處理技術，使後端圖像傳感器處理器(Image sensor processor, ISP)，不需要額外使用不同的處理器或者不同的調適參數。 且由於本文使用的影像傳感器硬體架構，能一次提供兩張長與短曝光的圖像，並減少了長與短曝光的時間間隔，使鬼影等現象不易產生。
接著改良現有的Tone mapping處理技術，分成兩部份處理，一部分考慮連續影像而非靜態影像處理，在即時顯示的連續影像模式時使用單點式全域色調映射(Global tone mapping)的方法。而另一部分，考慮處理靜態影像時，可使用複雜度較高但影像表現較好的局部區域偵測(Local tone mapping)方法，使用者在拍攝影像時可以較精準的掌握拍照角度、環境、光線等。甚至，使用較接近線性的Tone mapping處理可使即時圖像更接近影像處理後的色彩表現，並整個系統可以搭配現有的ISP processing flow，使HDR模式與一般LDR拍攝模式可使用相同的調適參數。

Over the years, the camera is become the part of every person’s life. From the beginning, the traditional film camera first came out, Kodak Instamatic, and then came the digital camera. Now most of the people even don’t carry camera anymore, instead using a smart phone, it can take not bad picture and automatically upload to the cloud, or post to the social network.
Smart phone brings us a lot of benefits, and also very convenient. But the flip side of that, it has to be small and easy to be carried, which is limit the pixel size and performance of imaging sensors. This directly affects the dynamic range of the image. When run into a high contrast scene which existed with sun light and the shadows. There are bright and dark details need to be recorded. High Dynamic Range (HDR) processing method is used for solving this kind of problems.
So far, it already has lots of kind of HDR techniques, like Exposure Fusion. Recently there are many related research. It has the better image quality, and less the artifact, but the higher computational complexity. And there is the traditional HDR Tone Mapping method. It has the lower computational complexity, but easy to caused the halo artifact.
This thesis is focused on the Tone Mapping technique. We can use the highly efficiency HDR processing method on the low cost CMOS image sensor, do the real time display. And take into account the work flow of the back-end Image Sensor Processor (ISP). Let the HDR image flow and normal LDR image flow can use the same parameter, such as Auto White Balance (AWB), Color Correction Matrix (CCM) and Gamma Correction, etc.
With new hardware architecture, it can provide two different exposures in the same frame. New hardware is also much reduced the time interval between two exposure images, so the ghosting effect will less occurred.
Then, we improve the existing HDR processing technique; consider the video streaming instead of static imaging. We used a pixel based Global Tone Mapping method to deal with the video streaming, and a highly computational complexity Local Tone Mapping method for the static images (snap shot), which can provide better image quality. So the photographer can take control the light and shooting angle more easily by viewing the real time HDR video streaming.

[1] P. E. Debevec and J. Malik. Recovering high dynamic range radiance maps from photographs. In SIGGRAPH ’97, T. Whitted, Ed. Pages: 369–378.
[2] R. Fattal, D. Lischinski and M. Werman. Gradient domain high dynamic range compression. ACM Trans. Graph. (special issue SIGGRAPH 2002) 21, pages: 249-256, 2002.
[3] S. Kavadias, B. Dierickx, D. Scheffer, A. Alaerts, D. Uwaerts, J. Bogaerts. A Logarithmic Response CMOS Image Sensor with On-Chip Calibration. IEEE Journal of Solid-State Circuits, Volume: 35, Pages: 1146-1152, August 2000.
[4] M. Vatteroni, D. Covi, A. Sartori. Linear–Logarithmic CMOS Pixel With Tunable Dynamic Range. IEEE Transactions on Electron Devices, Volume: 58, Pages: 1108-1115, 2011.
[5] W. F. Chou, S. F. Yeh, C. F. Chiu, C. C. Hsieh. A Linear-Logarithmic CMOS Image Sensor With Pixel-FPN Reduction and Tunable Response Curve. IEEE Sensors Journal, Volume: 14, Pages: 1625-1632, 2014.
[6] H. Zhao, B. Shi, C. Fernandez-Cull, S. K. Yeung, R. Raskar. Unbounded High Dynamic Range Photography Using a Modulo Camera. Computational Photography (ICCP), Pages: 1-10, 2015 IEEE.
[7] M. Tico; N. Gelfand and K. Pulli. Motion-blur-free Exposure Fusion. 2010 IEEE International Conference on Image Processing, Pages: 3321-3324, 2010.
[8] J. An, S. H. Lee, J. G. Kuk and N. I. Cho. A Multi-exposure Image Fusion Algorithm without Ghost Effect. Acoustics, Speech and Signal Processing (ICASSP), 2011 IEEE International Conference on. Pages: 1565-1568. 2011.
[9] T. Mertens, J. Kautz and F. V. Reeth. Exposure Fusion. Computer Graphics and Applications. 2007 PG ’07. 15th Pacific Conference on, Pages: 382-390, Oct. 29 2007-Nov.2.
[10] E. Reinhard, M. Stark, P. Shirley, J. Ferwerda. Photographic Tone Reproduction for Digital Images. Proceedings of SIGGRAPH 2002, Pages: 267-276. 2002.
[11] J. Duan, G. Qiu, and M. Chen. Comprehensive Fast Tone Mapping for High Dynamic Range Image Visualization. In Pacific Graphics Conference on Computer Graphics and Applications, 2005.
[12] F. Duran, J. Dorsey. Fast Bilateral Filtering for the Display of High-Dynamic-Range Images. ACM Trans. on Graphics 21 (2002) Proc. of SIGGRAPH conference. 2002.
[13] S. Paris, S. W. Hasinoff and J. Kautz. Local Laplacian Filters: Edge-aware Image Processing with a Laplacian Pyramid. ACM Transactions on Graphics (Proceedings of SIGGRAPH 2011), 2011.
[14] Y. Piao, W. Xu and Y. Piao. Method of Auto Multi-Exposure for High Dynamic Range Imaging. 2010 International Conference on Computer, Mechatronics, Control and Electronic Engineering, Pages: 93-97, 2010.
[15] K. F. Huang and J. C. Chiang. Intelligent Exposure Determination for High Quality HDR Image Generation. 2013 IEEE International Conference on Image Processing, Pages: 3201-3205, 2013.
[16] K. Kim and I. C. Park, Combined Image Signal Processing for CMOS Image Sensors. 2006 IEEE International Symposium on Circuits and Systems, Pages: 4 pp., 2006.
[17] W. Lee, J. Kim. A Cost-Effective Demosaicked Image Enhancement for a Single Chip CMOS Image Sensor. IEEE Worksop on Signal Processing Systems Design and Implementation, Pages: 148-153, 2005.
[18] Android: Camera HAL subsystem. 取自網路來源：http://source.android.com/devices/camera/camera3_requests_hal.html
[19] J. Nakamura. Image Sensors and Signal Processing for Digital Still Cameras. CRC Press, August 5, 2005.
[20] R. K. Thakur, A. Tripathy and A. K. Ray. A Design Framework of Digital Camera Images Using Edge Adaptive and Directionally Weighted Color Interpolation Algorithm. Computational Sciences and Optimization, 2009. CSO 2009. International Joint Conference on, Pages: 905-909, 2009.
[21] Pyramid (image processing). 取自網路來源：https://en.wikipedia.org/wiki/Pyramid_(image_processing)
[22] Solving the Discrete Poisson Equation using Jacobi, SOR, Conjugate Gradients, and the FFT Table of Contents, U.C. Berkeley, 1996, 取自網路來源：http://people.eecs.berkeley.edu/~demmel/cs267/lecture24/lecture24.html
[23] Solving the Discrete Poisson Equation using Multigrid, U.C. Berkeley, 1996, 取自網路來源：http://people.eecs.berkeley.edu/~demmel/cs267/lecture25/lecture25.html
[24] M. Saraniti, A. Rein, G. Zandler, P. Vogl, P. Lugli. An Efficient Multigrid Poisson Solver for Device Simulations. IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems. Vol.15, No.2, February, 1996.