Construction of Shadow Model by Robust Features to Illumination Changes

Construction of Shadow Model by Robust Features to Illumination Changes

Shuya Ishida, Shinji Fukui, Yuji Iwahori, M. K. Bhuyan, Robert J. Woodham
Copyright: © 2013 |Pages: 11
DOI: 10.4018/ijsi.2013100104
(Individual Articles)
No Current Special Offers


Methods in the field of computer vision need a shadow detection because shadows often have a harmful effect on a result. A new shadow detection method is proposed in this paper. The proposed method is based on the shadow model. The model is constructed by robust features to illumination changes. The proposed method uses the difference of chrominance (UV) components of luma chrominance (YUV) color space between the background image and the observed image, Normalized Vector Distance, Peripheral Increment Sign Correlation image and edge information. These features remove shadow effects in part. The proposed method can construct the effective shadow model by using the features. In addition, the result is improved by the region based method and the shadow model is updated. The proposed method can extract shadows accurately. Results are demonstrated by the experiments using the real videos.
Article Preview

Features Robust To Illumination Changes

Normalized Vector Distance

After dividing an image into small blocks, the ND at a pixel x is calculated by Equation 1:

(1) where ijsi.2013100104.m02 and ijsi.2013100104.m03 mean the irradiance data at ijsi.2013100104.m04 in the observed image and the background image respectively, ijsi.2013100104.m05 means an irradiance vector for a block of the observed image which includes ijsi.2013100104.m06 and ijsi.2013100104.m07 means that for the block of the background image.

The direction of ijsi.2013100104.m08 does not change much by the effect of illumination changes (Nagaya, Miyatake, Fujita, Ito & Ueda, 1996). ND is a robust feature to them and can remove the effect of shadows in part.

Complete Article List

Search this Journal:
Volume 12: 1 Issue (2024)
Volume 11: 1 Issue (2023)
Volume 10: 4 Issues (2022): 2 Released, 2 Forthcoming
Volume 9: 4 Issues (2021)
Volume 8: 4 Issues (2020)
Volume 7: 4 Issues (2019)
Volume 6: 4 Issues (2018)
Volume 5: 4 Issues (2017)
Volume 4: 4 Issues (2016)
Volume 3: 4 Issues (2015)
Volume 2: 4 Issues (2014)
Volume 1: 4 Issues (2013)
View Complete Journal Contents Listing