- 2k Downloads
Atmospheric conditions induced by suspended particles, such as fog and haze, severely alter the scene appearance. Restoring the true scene appearance from a single observation made in such bad weather conditions remains a challenging task due to the inherent ambiguity that arises in the image formation process. In this paper, we introduce a novel Bayesian probabilistic method that jointly estimates the scene albedo and depth from a single foggy image by fully leveraging their latent statistical structures. Our key idea is to model the image with a factorial Markov random field in which the scene albedo and depth are two statistically independent latent layers and to jointly estimate them. We show that we may exploit natural image and depth statistics as priors on these hidden layers and estimate the scene albedo and depth with a canonical expectation maximization algorithm with alternating minimization. We experimentally evaluate the effectiveness of our method on a number of synthetic and real foggy images. The results demonstrate that the method achieves accurate factorization even on challenging scenes for past methods that only constrain and estimate one of the latent variables.
KeywordsDefogging Dehazing Scene albedo Scene depth Markov random field Bayesian estimation
Unable to display preview. Download preview PDF.
- Fattal, R. (2008a) http://www.cs.huji.ac.il/~raananf/projects/defog/index.html.
- Fattal, R. (2008b). Single image dehazing. In Proc. of ACM SIGGRAPH (Vol. 27, pp. 1–9). Google Scholar
- Grewe, L., & Brooks, R. (1998). Atmospheric attenuation reduction through multi-sensor fusion. In Proc. of SPIE sensor fusion: architectures, algorithms and applications II (Vol. 3376, pp. 102–109). Google Scholar
- He, K., Sun, J., & Tang, X. (2009a). http://personal.ie.cuhk.edu.hk/~hkm007/cvpr09/.
- He, K., Sun, J., & Tang, X. (2009b). Single image haze removal using dark channel prior. In Proc. of IEEE int’l conf on comp. vision and pattern recognition (Vol. 1). Google Scholar
- Kim, J., & Zabih, R. (2002). Factorial Markov random fields. In Proc. of European conf on comp. vision (pp. 321–334). Google Scholar
- Komodakis, N., & Paragios, N. (2009). Beyond pairwise energies: efficient optimization for higher-order MRFs. In Proc. of IEEE int’l conf on comp. vision and pattern recognition (pp. 2985–2992). Google Scholar
- Kopf, J., Neubert, B., Chen, B., Cohen, M., Cohen-Or, D., Deussen, O., Uyttendaele, M., & Lischinski, D. (2008a). http://johanneskopf.de/publications/deep_photo/dehazing/index.html.
- Kratz, L., & Nishino, K. (2009). Factorizing scene albedo and depth from a single foggy image. In Proc. of IEEE int’l conf on comp. vision (pp. 1701–1708). Google Scholar
- McCartney, E. (1975). Optics of the atmosphere: scattering by molecules and particles. New York: Wiley. Google Scholar
- Narasimhan, S., & Nayar, S. (2003a). Shedding light on the weather. In Proc. of IEEE int’l conf on comp. vision and pattern recognition (pp. 665–672). Google Scholar
- Narasimhan, S. G., & Nayar, S. K. (2003c). Interactive (De)weathering of an image using physical models. In ICCV workshop on color and photometric methods in comp. vision (CPMCV). Google Scholar
- Narasimhan, S., Wand, C., & Nayar, S. (2002). All the images of an outdoor scene. In Proc. of European conf on comp. vision (pp. 148–162). Google Scholar
- Schechner, Y. (2003). http://webee.technion.ac.il/~yoav/research/instant-dehazing.html.
- Szeliski, R., Zabih, R., Scharstein, D., O Veksler, V. K., Agarwala, A., Tappen, M., & Rother, C. (2008). A comparative study of energy minimization methods for Markov random fields with smoothness-based priors. IEEE Transactions on Pattern Analysis and Machine Intelligence, 30(6), 1068–1080. CrossRefGoogle Scholar
- Tan, R. (2008a). http://people.cs.uu.nl/robby/fog/index.html.
- Tan, R. (2008b). Visibility in bad weather from a single image. In Proc. of IEEE int’l conf on comp. vision and pattern recognition (pp. 1–8). Google Scholar
- Tarel, J. P., & Hautière, N. (2009). Fast visibility restoration from a single color or gray level image. In Proc. of IEEE int’l conf on comp. vision. Google Scholar