隨著科技逐漸發達，日常生活中使用電腦視覺系統輔助的情形越漸普遍，以無人車為例，無人車上裝載著影像感測器，並使用影像感測器捕捉的畫面進行辨識與偵測，然而若行駛在一個有霧場景中，可能會因捕捉的畫面不清晰，導致辨識錯誤的情況發生，因此擁有一張清晰的場景影像來增加辨識正確率以降低事故發生的風險是非常重要的。
在做除霧處理前，必須得到一個有霧影像，並計算出整張影像因懸浮粒子而造成的介質傳輸率圖，以及大氣光所造成的亮度偏差，最後將這三個數值代入大氣散射物理模型中，獲得除霧後的影像。
本論文以卷積神經網路與FPGA(Field Programmable Gate Array)實現除霧系統，使用一個事先訓練好的卷積神經網路，當影像感測器在一個有霧的場景中接收影像資訊時，有霧資訊影像會先做維度的縮減，再進入到卷積網路中，生成霧霾的相關特徵，得到介質傳輸率圖，並藉由大氣散射物理模型，復原成一張去除霧之影像，最後將這整套系統實作在FPGA(Field Programmable Gate Array)板上，搭配相機鏡頭模組模擬實境情況，得到了一個完整的硬體除霧系統。

With the development of technology, computer vision system make our life more convenient. For example, the smart car use image sensor to recognize pedestrian or traffic signs, but the recognition error rate of the image sensor increases in bad weather condition, such as fog or haze. It is important to have a clear scene image to decrease recognition error rate and reduce the risk of accidents. In this study, we implement a dehaze system, based on atmospheric scattering model, on Field Programmable Gate Array (FPGA) using convolutional neural network(CNN). The key to achieve haze removal is to estimate a medium transmission map which indicating the light transmission rate under the medium for an input haze image. To reduce the computation time, the input image is down scaled to estimate its haze feature in order to obtain its corresponding medium transmission map. Then the estimated medium transmission map is up scaled to as the same size. The up scaled medium transmission map is used to remove the haze from the input image. To evaluate the effective of our work, both of the software version and FPGA version results are compared. It is shown that they are consistent. Our system is implemented on Altera DE2-115 with camera module, the frame rate is 5 fps under 100MHz clock rate.

[1] Arrigo Benedetti, Andrea Prati, Nello Scarabottolo. “Image convolution on FPGAs: the implementation of a multi-FPGA FIFO structure” , Euromicro Conference ,Vasteras, Sweden, 27 Aug. 1998, pp.123-130.
[2] A. Horé, D. Ziou, “Image quality metrics: Psnr vs. ssim”, IEEE International Conference on Pattern Recognition (ICPR), Istanbul, Turkey, 23-26 Aug. 2010, pp. 2366-2369.
[3] B. Bosi, G. Bois, and Y. Savaria, “Reconfigurable pipelined 2-D convolvers for fast digital signal processing”, IEEE Transactions on Very Large Scale Integration (VLSI) Systems, No. 3, Vol. 7, Sept. 1999, pp. 299-308.
[4] Bolun Cai, Xiangmin Xu; Kui Jia, Chunmei Qing, Dacheng Tao, “DehazeNet: An End-to-End System for Single Image Haze Removal”, IEEE Transactions on Image Processing, No. 11, Vol. 25, Nov. 2016, pp.5187-5198.
[5] Cl´ement Farabet, Cyril Poulet, Jefferson Y. Han, Yann LeCun, “CNP: An FPGA-based processor for Convolutional Networks”, Field Programmable Logic and Applications, Prague, Czech Republic, 31 Aug.-2 Sept. 2009, pp. 32-37.
[6] Chieh-Chi Kao, Jui-Hsin Lai, Shao-Yi Chien, “VLSI Architecture Design of Guided Filter for 30 Frames/s Full-HD Video”, IEEE Transactions on Circuits and Systems for Video Technology, No. 3, Vol. 24, Mar. 2014, pp.513-524.
[7] C. Dong, C. C. Loy, K. He, and X. Tang., “Learning a deep convolutional network for image super-resolution”, 2015 IEEE International Conference on Computer Vision Workshop (ICCVW), Santiago, Chile, 7-13 Dec. 2015, pp. 184-199.
[8] F. Liu, C. Shen, and G. Lin., “Deep convolutional neural fields for depth estimation from a single image”, Computer Vision and Pattern Recognition (CVPR), Boston, MA, USA, 7-12 June 2015, pp.5162-5170.
[9] I. Goodfellow, D. Warde-farley, M. Mirza, A. Courville, and Y. Bengio, “Maxout networks”, in Proceedings of the 30th International Conference on Machine Learning (ICML-13), 17-19 June 2013, Atlanta, Georgia, USA, pp. 1319-1327.
[10] Irfan Riaz, Teng Yu, Yawar Rehman, Hyunchul Shin, “Single image dehazing via reliability guided fusion”, Journal of Visual Communication and Image Representation, Vol.40, Part A, Jun. 2016, pp. 85-97.
[11] J. Kopf, B. Neubert, B. Chen, M. Cohen, D. Cohen-Or, O. Deussen, M. Uyttendaele, and D. Lischinski, “Deep photo: Model-based photograph enhancement and viewing”, in ACM Transactions on Graphics(TOG) , No. 5, Vol. 27, Jan. 2008, pp. 116-125.
[12] Jifeng Dai, Haozhi Qi, Yuwen Xiong, Yi Li, Guodong Zhang, Han Hu, Yichen Wei, “Deformable Convolutional Networks”, International Conference on Computer Vision (ICCV), Venice, Italy, 22-29 Oct. 2017.
[13] K. He, J. Sun, and X. Tang, “Single image haze removal using dark channel prior”, IEEE Trans. Pattern Anal. Mach. Intell. , No. 12, Vol. 33, Dec. 2011, pp. 2341-2353.
[14] Kaiming He, Jian Sun, Xiaoou Tang, “Guided Image Filtering”, IEEE Transaction on Pattern Analtsis and Machine Intelligence, No. 6, Vol. 35, Jun. 2013, pp.1397-1409.
[15] K. Tang, J. Yang, and J. Wang, “Investigating haze-relevant features in a learning framework for image dehazing”, in IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Columbus, OH, USA, 23-28 June, 2014, pp. 2995– 3002.
[16] Kaiming He, Jian Sun , “Fast guided filter”, Computer Vision and Pattern Recognition, Boston, MA, USA , 7-12 Jun, 2015.
[17] Magnus Halvorsen, “Hardware Acceleration of Convolutional Neural Networks”, Norwegian University of Science and Technology Department of Computer and Information Science, Master of Science in Computer Science, Jun. 2015.
[18] Q. Zhu, J. Mai, and L. Shao, “A fast single image haze removal algorithm using color attenuation prior”, IEEE Transactions on Image Processing, No. 11, Vol. 24, June 2015, pp. 3522-3533.
[19] R. G. Shoup, “Parameterized convolution filtering in a field programmable gate array,” Selected papers from the Oxford 1993 international workshop on field programmable logic and applications on More FPGAs. Oxford, United Kingdom, 1993, pp. 274–280.
[20] Robby T. Tan, “Visibility in Bad Weather from a Single Image”, in IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Anchorage, AK, USA, 23-28 June, 2008, pp. 1-8.
[21] Shree K. Nayar, Srinivasa G. Narasimhan, “Vision and the Atmosphere,” Internation Journal of Comeputer Vision, No. 3, Vol. 48, Aug. 2002, pp.233-254.
[22] S. G. Narasimhan and S. K. Nayar, “Contrast restoration of weather degraded images”, IEEE Transactions on Pattern Analysis and Machine Intelligence, No. 6, Vol. 25, June 2003, pp. 713-724.
[23] S. L. Chen, H. Y. Huang, and C. H. Luo, “A low-cost high-quality adaptive scalar for real-time multimedia applications”, IEEE Transactions on Circuits and Systems for Video Technology, No. 11, Vol. 21, Nov. 2011, pp. 1600-1611.
[24] Shih-Lun Chen, “VLSI Implementation of a Low-Cost High-Quality Image Scaling Processor”, IEEE Transactions on Circuits and Systems II: Express Briefs, No. 1, Vol. 60, Jan. 2013, pp.31-35.
[25] Shih-Lun Chen, “VLSI Implementation of an Adaptive Edge-Enhanced Image Scalar for Real-Time Multimedia Applications” , IEEE Transactions on Circuits and Systems for Video Technology, No. 9 , Vol. 23, Sept. 2013, pp.1510-1522.
[26] V. Nair and G. E. Hinton, “Rectified linear units improve restricted boltzmann machines” , in Proceedings of the 27th International Conference on Machine Learning (ICML-10), Haifa, Israel, June 21-24, 2010, pp. 807-814.
[27] Y. LeCun, L. Bottou, Y. Bengio, and P. Haffner, “Gradient-based learning applied to document recognition”, Proceedings of the IEEE, No. 11, Vol. 86, Nov. 1998, pp. 2278-2324.
[28] Y. Y. Schechner, S. G. Narasimhan, and S. K. Nayar, “Instant dehazing of images using polarization”, in IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Kauai, HI, USA, USA, 8-14 Dec. 2001, pp. 325-332.
[29] Yongmei Zhou, Jingfei Jiang, “An FPGA-based Accelerator Implementation for Deep Convolutional Neural Networks”, 2015 4th International Conference on Computer Science and Network Technology (ICCSNT 2015), Harbin, China, 19-20 Dec. 2015, pp.829-832.
[30] Zhou Wang, Alan Conrad Bovik, Hamid Rahim Sheikh, Eero P. Simoncelli, “Image Quality Assessment: From Error Visibility to Structural Similarity”, IEEE Transaction on Image Processing, No. 4, Vol. 13, Apr. 2004, pp.600-612.
[31] Z. Cai, Q. Fan, R. Feris, and N. Vasconcelos. "A unified multi-scale deep convolutional neural network for fast object detection", In ECCV, Amsterdam, The Netherlands, 8-16 Oct. 2016.