在日常生活中，陰影隨處可見。人類視覺系統可以輕易分辨陰影，但若要電
腦偵測出一張影像中的陰影，並將影像中的陰影移除是非常困難的。此外，在影
像處理領域，雖然陰影可以幫助我們了解影像場景架構，但當影像在做其他應
用，例如物件追蹤、物件辨識或影像切割時，陰影卻會影響結果。因此陰影偵
測(Shadow detection)與陰影移除(Shadow removal)便成為影像處理中一個重要的 課題。
現今已有許多偵測與移除陰影的技術，但陰影邊緣半影(Penumbra)區域卻很 難正確的偵測。因此本論文針對陰影邊緣提出三個方法，第一，若影像紋理簡
單，則用影像修補方式修補邊緣。第二，則對邊緣做影像超解析，並二次偵測陰
影。第三，使用半影的移除方法結合骨架化達到移除自動化。如此一來，陰影邊
緣便能更準確的自動移除。

Shadow can be seen everywhere in our daily life. The visual system of human
beings can distinguish shadow easily, but it is difficult for a computer to detect
and remove the shadow in an image. In addition, in the image processing field,
shadow can help us understand the architecture of a scene and the direction of a
light source, but in some applications, such as object tracking, object recognition
and image segmentation, shadow may cause failure. Therefore, shadow detection
and shadow removal become important issues in image processing.
Nowadays, there are many techniques for detecting and removing shadows, but
it is still hard to detect penumbra accurately. Therefore, this paper proposed three
methods for detecting a shadow boundary. First, if the texture of an image is simple,
we use inpainting to inpaint the boundaries. Second, if it is complex, we get the
blocks along the shadow boundary, perform super-resolution for the blocks, and
detect the shadow again. Last, we can use shadow removal for penumbra. As a
result, we can remove penumbra more correctly.

[1] Ariel Amato, Ivan Huerta, Mikhail G Mozerov, F Xavier Roca, and Jordi Gonzalez.
Moving cast shadows detection methods for video surveillance applications.
In Wide Area Surveillance, pages 23–47. Springer, 2014.
[2] Eli Arbel and Hagit Hel-Or. Shadow removal using intensity surfaces and texture
anchor points. IEEE Trans. Pattern Anal. Mach. Intell., 33(6):1202–1216,
June 2011.
[3] Antonio Criminisi, Patrick Perez, and Kentaro Toyama. Region filling and
object removal by exemplar-based image inpainting. IEEE Transactions on
Image Processing, 13/9:1200–1212, September 2004.
[4] Kaushik Deb and Ashraful Huq Suny. Shadow detection and removal based on
ycbcr color space. Smart CR, 4:23–33, 2014.
[5] D Ebert, P Brunet, and I Navazo. An augmented fast marching method for
computing skeletons and centerlines. 2002.
[6] Eva Eibenberger and Elli Angelopoulou. The narrow-band assumption in logchromaticity
space. In ECCV Workshops (1), pages 76–89, 2010.
[7] Graham Finlayson, Mark Drew, and Cheng Lu. Intrinsic images by entropy
minimization. Computer Vision-ECCV 2004, pages 582–595, 2004.
[8] Han Gong and Darren Cosker. Interactive shadow removal and ground truth
for variable scene categories. In Proceedings of the British Machine Vision
Conference. BMVA Press, 2014.
[9] R. Guo, Q. Dai, and D. Hoiem. Single-image shadow detection and removal
using paired regions. In CVPR 2011, pages 2033–2040, June 2011.
[10] Kwang Kim and Younghee Kwon. Example-based learning for single-image
super-resolution. Pattern Recognition, pages 456–465, 2008.
[11] Feng Liu and Michael Gleicher. Texture-consistent shadow removal. Computer
Vision–ECCV 2008, pages 437–450, 2008.
[12] D. Rufenacht, C. Fredembach, and S. Susstrunk. Automatic and accurate
shadow detection using near-infrared information. IEEE Transactions on Pattern
Analysis and Machine Intelligence, 36(8):1672–1678, Aug 2014.
[13] Ashraful Huq Suny and Nasrin Hakim Mithila. A shadow detection and removal
from a single image using lab color space. IJCSI International Journal of
Computer Science Issues, 10(4), 2013.
[14] T. F. Yago Vicente, M. Hoai, and D. Samaras. Leave-one-out kernel optimization
for shadow detection and removal. IEEE Transactions on Pattern Analysis
and Machine Intelligence, PP(99):1–1, 2017.