隨著深度學習發展的日益增長，現今深度學習已成為火紅的議題，各式圖像與視覺的處理後標榜深度學習的商品及服務搶攻市場占有率。而深度學習之中，除了資料數量的需求龐大，對於學習的輸入圖像有著很大的品質要求。因此各式數位影像處理技術應運而生，諸如邊緣填充、模糊化[1]、平滑化、銳利化、影像濾波…等等技術。但在這些技術大放異彩的同時，應用於IOT小裝置上時，運算處理效率的議題也隨之浮現，在攜帶性裝置與IOT越做越微小的世代中，如何在兼具相同運算效果的同時兼具快速處理的運算方法已成為不可或缺的能力。
無人駕駛與電腦視覺等技術正在逐漸一步一步塑造出我們對未來日常生活的構想，其中人工智慧與電腦視覺等資源消耗高的問題亦成為熱門話題，而由此衍伸出的分散式的同餘系統之核心函數的特性運算減少整體對於影像處裡或是深度學習的執行時間或是訓練速度的運算並利用同餘系統之性質，保證資訊在分散式處理後的正確性與安全性。因此在本論文中，我們打造一個系統，該系統能將未經處理的影像透過同餘系統的特性，藉由改良演算法使原先數位影像處理中繁瑣的運算速度獲得50% 提升。

Ever since the ever-growing Deep-Learning development, this technology has been gaining popularity. Countless amounts of image and visual processing products and services advertising Deep-Learning have taken the market globally by storm. Adequate amount of data is required to perform Deep-Learning as well as the quality of data. Therefore, numerous digital processing technologies have been created to meet the needs, such as buffering, blurring, smoothing, sharpening, filtering… and so on. However, when implementing IOT and tiny devices, the problem of computing efficiency arises. In an era when IOT and portable devices became more and more tiny, how to preserve efficiency while shrinking the size is a must-have ability.
Unmanned vehicle and computer vision techniques are gradually shaping the future; thus, the high consumption of resource artificial intelligence requires becomes a trend. And Core Function of Residue Number System is capable of lowering the runtime or training speed of image processing by nature. This allows the such algorithms to be performed on micro devices while also providing confidentiality. We are able to build a system which is capable of boosting image process operating speed by 50% by modifying the algorithms and modifying the input image data.

[1] T. Indriyani, M. I. Utoyo, and R. Rulaniingtyas. “Comparison of Image Smoothing Methods on Potholes Road Images,” Journal of Physics: Confference Series, vol.1477, 2020.
[2] Gonnella, J. (1991). The application of core functions to residue number systems. IEEE Transactions on Signal Processing, 39(1), 69–75. doi:10.1109/78.80766
[3] Kong, Y., Asif, S., & Khan, M. A. U. (2015). Modular multiplication using the core function in the residue number system. Applicable Algebra in Engineering, Communication and Computing, 27(1), 1–16. doi:10.1007/s00200-015-0268-1
[4] Tenca, A. F., & Koc, C. K. (2003). A scalable architecture for modular multiplication based on montgomery’s algorithm. IEEE Transactions on Computers, 52(9), 1215–1221. doi:10.1109/tc.2003.1228516
[5] Garner, H. L. (1959). The residue number system. Papers Presented at the the March 3-5, 1959, Western Joint Computer Conference on XX - IRE-AIEE-ACM ’59 (Western). doi:10.1145/1457838.1457864
[6] Tailor, A. (1986). Introductory digital image processing: a remote sensing perspective. Image and Vision Computing, 4(4), 229. doi:10.1016/0262-8856(86)90052-1
[7] Marr, D., & Hildreth, E. (1980). Theory of Edge Detection. Proceedings of the Royal Society B:Biological Sciences, 207(1167), 187–217. doi:10.1098/ rspb.1980.0020
[8] Miller, D. D. ,Polky, J. N. , King,J.R.: A survey of soviet developments in residue number theory applied to digital filtering. In: Proceedings off the 26th Midwest Symposium on Cirrcuits and Systems (1983)
[9] Thu Van Vu. (1985). Efficient Implementations of the Chinese Remainder Theorem for Sign Detection and Residue Decoding. IEEE Transactions on Computers, C-34(7), 646–651. doi:10.1109/tc.1985.1676602
[10] Burgess, N. (n.d.). Scaling an RNS number using the core function. 16th IEEE Symposium on Computer Arithmetic, 2003. Proceedings. doi:10.1109/arith.2003.1207687
[11] David Duce, Ivan Herman, and Bob Hopgood, “Web 2D Graphics File Formats”, COMPUTER GRAPHICS forum, Vol. 21, Issue 1, pp. 43-64, March 2002
[12] Lin Hong, Yifei Wan, & Jain, A. (1998). Fingerprint image enhancement: algorithm and performance evaluation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 20(8), 777–789. doi:10.1109/34.709565
[13] Milanfar, P. (2013). A Tour of Modern Image Filtering: New Insights and Methods, Both Practical and Theoretical. IEEE Signal Processing Magazine, 30(1), 106–128. doi:10.1109/msp.2011.2179329
[14] Yu, Y., Yang, N., Yang, C., … Nyima, T. (2019). Memristor bridge-based low pass filter for image processing. Journal of Systems Engineering and Electronics, 30(03), 448–455. doi:10.21629/jsee.2019.03.02
[15] Metwalli, M. R., Nasr, A. H., Farag Allah, O. S., & El-Rabaie, S. (2009). Image fusion based on principal component analysis and high-pass filter. 2009 International Conference on Computer Engineering & Systems. doi:10.1109/icces.2009.5383308
[16] Li, Q., Wang, Q., & Li, X. (2021). Exploring the Relationship Between 2D/3D Convolution for Hyperspectral Image Super-Resolution. IEEE Transactions on Geoscience and Remote Sensing, 1–11. doi:10.1109/tgrs.2020.3047363
[17] Clark, J. J. (1989). Authenticating edges produced by zero-crossing algorithms. IEEE Transactions on Pattern Analysis and Machine Intelligence, 11(1), 43–57. doi:10.1109/34.23112
[18] Kong, H., Akakin, H. C., & Sarma, S. E. (2013). A Generalized Laplacian of Gaussian Filter for Blob Detection and Its Applications. IEEE Transactions on Cybernetics, 43(6), 1719–1733. doi:10.1109/tsmcb.2012.2228639
[19] Shen, D., Chiu, C., & Huang, P. (2006). Modified Laplacian Filter and Intensity Correction Technique for Image Resolution Enhancement. 2006 IEEE International Conference on Multimedia and Expo. doi:10.1109/icme.2006.262571
[20] Nguyen, A.-D., Choi, S., Kim, W., Ahn, S., Kim, J., & Lee, S. (2019). Distribution Padding in Convolutional Neural Networks. 2019 IEEE International Conference on Image Processing (ICIP). doi:10.1109/icip.2019.8803537
[21] Xinwen Fu, Graham, B., Bettati, R., & Wei Zhao. (n.d.). On effectiveness of link padding for statistical traffic analysis attacks. 23rd International Conference on Distributed Computing Systems, 2003. Proceedings. doi:10.1109/icdcs.2003.1203483
[22] Ito, I. (2014). A computing method of double linear correlation for mirror image matching. 2014 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). doi:10.1109/icassp.2014.6854114