近年來AI電腦視覺與醫學技術逐漸結合，開創各種更方便、精確的工具協助醫療。Canfield醫美儀器透過特殊波長分析黑色素與血紅蛋白的功能，使醫生更容易跟病人解釋治療黑斑前後的差別，進而避免掉許多醫療糾紛。本研究提出「基於Pix2PixHD之色素性臉部皮膚分析與診斷建議系統」可與目前使用之醫美儀器Canfield媲美，本研究具有下列特色。 (1)基於Pix2PixHD臉部皮膚色素分布分析技術：目前尚無透過影像直接分析黑色素與血紅蛋白的技術，此技術可不用特殊儀器分析皮膚狀況，不僅價低成本也提高攜帶性。分析結果以人眼感受來判斷與Canfield分析結果相似度已高達90%以上；(2)臉部黑斑分析技術：以往黑斑診斷都是以醫生主觀判斷，自動分析黑斑的技術較少。透過這個技術找出黑斑區域及分析嚴重度。透過偵測黑斑與血紅蛋白有相關的區域，找出需要特別治療的區域。針對黑斑症狀的病人黑斑偵測結果可達90%以上;(3)目前色素性皮膚相關資料庫較少。本團隊與黃柏翰皮膚科診所合作提出臉部皮膚色素分布資料庫，分成訓練資料集3000組及測試資料集678組。每組有標準白光、偏振光、黑色素分、血紅蛋白分布4種3312*3808大小照片。

In recent years, artificial intelligence(AI) computer vision and medical technology have gradually combined to create a variety of more convenient and accurate tools to assist medical treatment. The Canfield, medical beauty instrument, has the functions of melanin and hemoglobin through special wavelengths which make doctors explain difference before and after treatment of melasma more easily and avoid many medical disputes. This research proposes "Pix2PixHD-based pigmented facial skin analysis and diagnosis suggestion system" is comparable to Canfield, the currently used medical beauty instrument. This research has the following characteristics. (1) Based on Pix2PixHD facial skin pigment distribution analysis technology: At present, there is no technology to directly analyze melanin and hemoglobin through images. This technology can analyze the skin condition without specific equipment, and it is not only cost-effective but also improves portability. The similarity with Canfield analysis result is as high as 90% judged by human; (2) Facial melasma analysis technology: In the past, melasma diagnosis was based on the subjective of doctors, and the automatic analysis technology of melasma was less. Through this technology to find out the region and severity of melasma. By detecting the region related to melasma and hemoglobin, identify the region that require special treatment. For patients who have melasma symptoms, the results of melasma detection can reach more than 90% judged by human; (3) Pigmented facial skin dataset: There are currently few datasets related to pigmented skin. Our team cooperated with HUANGDERM to propose a dataset of facial skin pigment distribution. The dataset is divided into 3000 groups of training data sets and 678 groups of test data sets. Each sets have 4 categories including standard white light, polarized light, melanin, and hemoglobin distribution. The resolution of each image in the dataset is 3312*3808.

[1]R. Bhagwat, "The relationship of skin tone to physical and mental health outcomes in South Asian Americans," Rutgers University-Graduate School-New Brunswick, 2012.
[2]N. G. Jablonski, "The evolution of human skin and skin color," Annu. Rev. Anthropol, vol. 33, pp. 585-623, 2004.
[3]H. Y. Kang and J.-P. Ortonne, "What should be considered in treatment of melasma," Annals of dermatology, vol. 22, no. 4, p. 373, 2010.
[4]R. Demirli, P. Otto, R. Viswanathan, S. Patwardhan, and J. Larkey, "RBX® technology overview," Canfield Systems White Paper, 2007.
[5]T. Chang and C. Huang, "Skin Condition Detection Based on Image Processing Techniques," in Proceeding of Conference on Information Technology and Applications in Outlying Island, 2010.
[6]S. E. Umbaugh, R. H. Moss, and W. V. Stoecker, "Automatic color segmentation of images with application to detection of variegated coloring in skin tumors," IEEE Engineering in Medicine and Biology Magazine, vol. 8, no. 4, pp. 43-50, 1989.
[7]Z. Wang and R. Li, "Facial Pore Detection Based on Characteristics of Skin Pigment Distribution," in IEEE International Conference on Image Processing (ICIP), 2019: IEEE, pp. 4439-4443.
[8]T. N. Tran, K. Drab, and M. Daszykowski, "Revised DBSCAN algorithm to cluster data with dense adjacent clusters," Chemometrics and Intelligent Laboratory Systems, vol. 120, pp. 92-96, 2013.
[9]X. Liu, J. Sun, and X. Wang, "Facial Spot Contour Extraction based on Color Image Processing," in International Conference on Biomedical Signal and Image Processing (ICBIP), 2019, pp. 87-92.
[10]C. Zhao, G.-z. Li, F. Li, Z. Wang, and C. Liu, "Qualitative and quantitative analysis for facial complexion in traditional Chinese medicine," BioMed Research International, vol. 2014, 2014.
[11]T. Mendonça, P. M. Ferreira, J. S. Marques, A. R. Marcal, and J. Rozeira, "PH 2-A dermoscopic image database for research and benchmarking," in IEEE engineering in medicine and biology society (EMBC), 2013: IEEE, pp. 5437-5440.
[12]P. Tschandl, C. Rosendahl, and H. Kittler, "The HAM10000 dataset, a large collection of multi-source dermatoscopic images of common pigmented skin lesions," Scientific data, vol. 5, no. 1, pp. 1-9, 2018.
[13]M. Combalia et al., "BCN20000: Dermoscopic lesions in the wild," arXiv preprint arXiv:1908.02288, 2019.
[14]B. C. Ennehar, O. Brahim, and T. Hicham, "An appropriate color space to improve human skin detection," INFOCOMP Journal of Computer Science, vol. 9, no. 4, pp. 1-10, 2010.
[15]G. B. Solanke and S. Gore, "GPU Accelerated Computing for Human Skin Colour Detection Using YCbCr Colour Model," in International Conference on Computing, Communication, Control and Automation (ICCUBEA), 2017: IEEE, pp. 1-6.
[16]K. B. Shaik, P. Ganesan, V. Kalist, B. Sathish, and J. M. M. Jenitha, "Comparative study of skin color detection and segmentation in HSV and YCbCr color space," Procedia Computer Science, vol. 57, pp. 41-48, 2015.
[17]R. F. Rahmat, T. Chairunnisa, D. Gunawan, and O. S. Sitompul, "Skin color segmentation using multi-color space threshold," in International Conference on Computer and Information Sciences (ICCOINS), 2016: IEEE, pp. 391-396.
[18]Y. Kim, I. Hwang, and N. I. Cho, "Convolutional neural networks and training strategies for skin detection," in IEEE International Conference on Image Processing (ICIP), 2017: IEEE, pp. 3919-3923.
[19]H. Zuo, H. Fan, E. Blasch, and H. Ling, "Combining convolutional and recurrent neural networks for human skin detection," IEEE Signal Processing Letters, vol. 24, no. 3, pp. 289-293, 2017.
[20]J. P. B. Casati, D. R. Moraes, and E. L. L. Rodrigues, "SFA: A human skin image database based on FERET and AR facial images," in IX workshop de Visao Computational, Rio de Janeiro, 2013.
[21]A. Martinez and R. Benavente, "The AR Face Database: CVC Technical Report, 24," 1998.
[22]P. J. Phillips, H. Moon, S. A. Rizvi, and P. J. Rauss, "The FERET evaluation methodology for face-recognition algorithms," IEEE Transactions on pattern analysis and machine intelligence, vol. 22, no. 10, pp. 1090-1104, 2000.
[23]S. L. Phung, A. Bouzerdoum, and D. Chai, "Skin segmentation using color pixel classification: analysis and comparison," IEEE transactions on pattern analysis and machine intelligence, vol. 27, no. 1, pp. 148-154, 2005.
[24]M. Kawulok, J. Kawulok, and J. Nalepa, "Spatial-based skin detection using discriminative skin-presence features," Pattern Recognition Letters, vol. 41, pp. 3-13, 2014.
[25]I. J. Goodfellow et al., "Generative adversarial networks," arXiv preprint arXiv:1406.2661, 2014.
[26]M. Mirza and S. Osindero, "Conditional generative adversarial nets," arXiv preprint arXiv:1411.1784, 2014.
[27]P. Isola, J.-Y. Zhu, T. Zhou, and A. A. Efros, "Image-to-image translation with conditional adversarial networks," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2017, pp. 1125-1134.
[28]J.-Y. Zhu, T. Park, P. Isola, and A. A. Efros, "Unpaired image-to-image translation using cycle-consistent adversarial networks," in Proceedings of the IEEE international conference on computer vision, 2017, pp. 2223-2232.
[29]M. Cordts et al., "The cityscapes dataset for semantic urban scene understanding," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 3213-3223.
[30]O. Ronneberger, P. Fischer, and T. Brox, "U-net: Convolutional networks for biomedical image segmentation," in International Conference on Medical image computing and computer-assisted intervention, 2015: Springer, pp. 234-241.
[31]T.-C. Wang, M.-Y. Liu, J.-Y. Zhu, A. Tao, J. Kautz, and B. Catanzaro, "High-resolution image synthesis and semantic manipulation with conditional gans," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 8798-8807.
[32]J. Johnson, A. Alahi, and L. Fei-Fei, "Perceptual losses for real-time style transfer and super-resolution," in European conference on computer vision, 2016: Springer, pp. 694-711.
[33]K. He, X. Zhang, S. Ren, and J. Sun, "Deep residual learning for image recognition," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 770-778.
[34]V. Dumoulin and F. Visin, "A guide to convolution arithmetic for deep learning," arXiv preprint arXiv:1603.07285, 2016.
[35]Y. Jia et al., "Multipath ghost suppression based on generative adversarial nets in through-wall radar imaging," Electronics, vol. 8, no. 6, p. 626, 2019.
[36]X. Mao, Q. Li, H. Xie, R. Y. Lau, Z. Wang, and S. Paul Smolley, "Least squares generative adversarial networks," in Proceedings of the IEEE international conference on computer vision, 2017, pp. 2794-2802.
[37]A. G. Pandya et al., "Reliability assessment and validation of the Melasma Area and Severity Index (MASI) and a new modified MASI scoring method," Journal of the American Academy of Dermatology, vol. 64, no. 1, pp. 78-83. e2, 2011.
[38]D. King. Dlib library. Available: http://dlib.net/.
[39]81 Facial Landmarks Shape Predictor. Available: https://github.com/codeniko/shape_predictor_81_face_landmarks.