Background segmentation in multicolored illumination environments
- PDF / 3,996,661 Bytes
- 13 Pages / 595.276 x 790.866 pts Page_size
- 91 Downloads / 210 Views
ORIGINAL ARTICLE
Background segmentation in multicolored illumination environments Nikolas Ladas1
· Paris Kaimakis2 · Yiorgos Chrysanthou1
Accepted: 11 September 2020 © Springer-Verlag GmbH Germany, part of Springer Nature 2020
Abstract We present an algorithm for the segmentation of images into background and foreground regions. The proposed algorithm utilizes a physically based formulation of scene appearance which explicitly models the formation of shadows originating from color light sources. This formulation enables a probabilistic model to distinguish between shadows and foreground objects in challenging images. A key component of the proposed method is an algorithm for estimating the illumination arriving at the scene. We evaluate our algorithm using synthetic and real-world data and show that the proposed method performs favorably against other commonly used segmentation methods. Keywords Segmentation · Background modeling · Shadow detection · Visibility decomposition
1 Introduction Image segmentation is an integral part of many computer vision applications such as security, navigation, autonomous driving and virtual/augmented reality. These applications require high segmentation accuracy at low computational cost. Image segmentation is an extensively studied area and a number of highly accurate methods exist, such as [17] which utilities Convolutional Neural Networks. However, when considering pixel-level methods, which are commonly used in resource-constrained devices, there is still room for improvement. This is especially true when considering scenes with uneven illumination such as shadows, directed lighting (such as spotlights) and multicolored light sources. These can be found in theaters, stadiums, art installations and movie production settings and often result in misclassifications. This work introduces a segmentation algorithm that maintains high segmentation accuracy in scenes with uneven illu-
B
Nikolas Ladas [email protected] Paris Kaimakis [email protected] Yiorgos Chrysanthou [email protected]
1
Department of Computer Science, University of Cyprus, Nicosia, Cyprus
2
Department of Computing, University of Central Lancanshire Cyprus, Larnaca, Cyprus
mination. Our algorithm operates on the pixel level of a video stream and utilizes a physically based formulation of scene appearance based on Lambertian reflectance. By explicitly modeling the formation of shadows, we avoid misclassifications which are commonly problematic for pixel-level segmentation methods. Our formulation drives a probabilistic model that enhances robustness and delivers the final segmentation. Our method can utilize an environment map, if available, which captures the scene’s incoming illumination, in order to boost segmentation accuracy. An example segmentation result can be seen in Fig. 1. To summarize, this work contributes the following:
1. We introduce a formulation for the purpose of image segmentation based on Lambertian reflectance. This formulation naturally models the formation of shadows and c
Data Loading...
