陰影去除之應用行之有年，其在眾多領域，諸如：車禍影像之辨識、空拍影
像增強、場景重建等等，皆有發展。手持式彩色3D 掃描容易遇到光源受限，導致
材質陰影無法消除的問題。過去關於3D 掃描的研究，多半著重於掃描形狀之精準
度，較少探討光影消除的問題，本論文便是希望能改善此情況而發想。
本論文致力於解決 3D 模型之光影問題，從而達到修正3D 模型表面顏色之效
果。本論文分為三個實驗，實驗一為分群聚類法分色，實驗二為光源估測校正，
實驗三為光影估測校正，方法不同，但目的皆為修正三3D 模型光影不均的色彩資
訊。
實驗一目的為透過分群聚類法，將3D 模型物體色分為目標群集後，使用群集
中之鮮豔色彩將其替代；實驗二則為透過一測量球估測光源分布，製作光分布對
照表後，據此將實驗模型的表面色彩校正；實驗三則為透過模擬光源與模型，對
其做遮蔽陰影運算後，利用Phong 光照模型的光影推測，消除3D 模型之光影問題。
研究最後比較 3D 模型上之同色區塊校正前後的色差與明度差。實驗一的方法
普遍效果不佳。實驗二之色差與明度改善率平均值分別為46.8%與58.9%，最高可
達75.1%與96.3%。實驗三之之色差與明度改善率平均值分別為41.2%與52.7%色
差改善率，最高可達67.3%與87.6%。雖然整體上不如實驗二的方法，但可以降低
某些陰影區色差，是值得進一步研究的方向。

Shadow removal technologies have been used in many different aspects such as
image recognition of car accidents, image enhancement of remote sensing, 3D scene
reconstruction, et al. In hand-held color 3D scanning, lighting directions normally are
not controllable. The shadow of the resulted 3D model therefore cannot be removed
easily. As previous 3D scanning studies mainly focus on spatial accuracy of the 3D
model, only few attempts on its shadow removal.
To eliminate the illumination of the color 3D scan models, three types of methods
were tested in this study. Experiment 1 used color clustering, Experiment 2
compensated the shadows by means of illumination estimation, and Experiment 3
compensated the shadows via illumination and shadow estimations. The methods used
in Experiment 1 first classifies colors of a 3D model based on color differences and
spatial locations. Afterward, the colors of each cluster are replaced by the most saturate
color in the cluster. Experiment 2 first scan a white ball to estimate the light distribution,
and then eliminate the shadows by normalizing the light distribution. Experiment 3 also
estimate the light distribution first, but it applies Phong reflection model to estimate the
shadow and use the results to eliminate the lighting effects. The results of shadow
removal of a color 3D scan model were estimated by the color differences between two
regions toward different directions but painting same color in the original model. The
results show the method used in Experiment 2 is more reliable.

[1] 3D Systems, Inc. (1988) “StereoLithography Interface Specification”.
[2] P. Bourke (2009) “Ply-polygon file format”. http://paulbourke.net/dataformats/ply.
[3] Wavefront Technologies. (n.d.) “Object Files (.obj)”. Advanced Visualizer Manual.
[4] 鄭正元、江卓培、林宗翰、林榮信、蘇威年、汪家昌、蔡明忠、賴維祥、鄭
逸琳、洪基彬編著 (2017) “3D 列印基層製造技術與應用”. 台北：全華圖書
股份有限公司, 3-6~7,18~19。
[5] B. T. Phong (1975) “ Illumination for computer generated pictures”.
Communications of ACM, 18(6):311–317.
[6] B. Smith (2006) “Illustration of the components of the Phong reflection
model (Ambient, Diffuse and Specular reflection)”.
https://upload.wikimedia.org/wikipedia/commons/6/6b/Phong_components_versio
n_4.png.
[7] 黃日鋒、詹文鑫、陳鴻興、胡國瑞、徐道義、孫沛立、羅梅君著(2011). “顯示
色彩工程學”. 台北：全華圖書股份有限公司, 第139 頁。
[8] M. Stokes, M. Anderson, S. Chandrasekar, and R. Motta (1996). “A Standard
Default Color Space for the Internet – sRGB”.
https://www.w3.org/Graphics/Color/sRGB
[9] J. Kender (1976). “Saturation, hue and normalized color”. Carnegie Mellon
University, Computer Science Dept., Pittsburgh, PA.
[10] A.K. Jain, M.N. Murty and P.J. Flynn (1999). “Data Clustering: A Review”. ACM
Computing Surveys, 31(3):264-323
[11] P. Berkhin (2002). “Survey of Clustering Data Mining Techniques”. Technical
Report, San Jose: Accrue Software.
[12] J.A. Hartigan. (1975). “Clustering algorithms”. John Wiley and Sons, Inc.
[13] J. A. Hartigan, M. A. Wong (1979). “Algorithm AS 136: A k-Means Clustering
Algorithm”. Journal of the Royal Statistical Society, Series C (Applied Statistics),
28(1):100–108.
[14] K. Fukunaga and L.D. Hostetler (1975). “The Estimation of the Gradient of a
Density Function, with Applications in Pattern Recognition”. IEEE Transactions
on Information Theory, 21(1):32–40.
[15] Y. Cheng (1995). “Mean Shift, Mode Seeking, and Clustering”. IEEE Transactions
on Pattern Analysis and Machine Intelligence, 17(8):790–799.
[16] C. Tomasi and R. Manduchi (1998). “Bilateral Filtering for Gray and Color
Images”. In Proceedings of the Sixth Internatmal Conference on Computer Vision,
839-846.
[17] P. Soille (1999). “Morphological Image Analysis; Principles and Applications”.
Berlin, Germany: Springer Press. (289-317)
[18] K. Schlüns and A. Koschan (2000). “Global and Local Highlight Analysis in Color
Images”, 1st International Conference on Color in Graphics and Image Processing
CGIP´2000. 151 Proc.
[19] Y. Xiao, E. Tsougenis and C. Tang (2014) “Shadow Removal from Single RGB-D
Images”. The Hong Kong University of Science and Technology.
[20] Q. Yang, K. Tan and N. Ahuja (2012) “Shadow Removal Using Bilateral Filtering”.
IEEE Transactions on Image Processing, 21(10).
[21] R. Guo, Q. Dai and D. Hoiem (2011) “Single-Image Shadow Detection and
Removal using Paired Regions”. Conference on Computer Vision and Pattern
Recognition.
[22] 賴正才、廖珗洲、黃培安、謝弦祐(2012) “視訊上之陰影消除技術的實現”. 資
訊科技國際研討會暨第二屆網路智能與應用研討會.
[23] M. Dellepiane, L. Benedetti and R. Scopigno (2010) “Removing shadows for color
projection using sun position estimation” . The 11th International Symposium on
Virtual Reality, Archaeology and Cultural Heritage VAST. 55–62
[24] J. Dorsey, H. Rushmeier and F. Sillion (2008), Digital Modeling of Material
Appearance, Morgan Kaufmann Publishers, 61-91.
[25] 詹凱琳 (2016) “三維掃描影像色彩修正技術”. 色彩與照明科技研究所碩士學
位論文.