Change detection robust to sudden illumination changes
Description of the method
In  a novel approach to deal with the problem of accurately segment the foreground from the background under the presence of heavy illuminations changes was proposed. Along the first stage of the proposed three-stage algorithm, the proposed robust measure is employed to perform robust visual correspondence so to extract a subset of pixels of the current frame which reliably belong to the current background. This is useful for successive stages of the algorithms in order to perform a tonal alignment of the background model to the current frame so to yield robustness and segmentation accuracy. More details can be found in .
Experimental results shown in the figure below and obtained using a sequence affected by real sudden illumination changes as well as a syntethic sequence affected by artificial brightness distortions show the efficacy of the proposed approach in accurately segment the foreground from the background even in presence of such conditions. Moreover, you can see the results of the proposed algorithm applied over a video sequence which was obtained from the MUSCLE website. Additional experimental results can be found in 
We provide an implementation of the algorithm proposed in  in the form of a pre-compiled library. The library can be used with any 32-bit or 64-bit VisualStudio projects. It can be freely used for research and dissemination purposes. If you use this software for any kind of publication or work, we kindly request to cite .
In the provided zip file, you will find, additionally to the library files, a test file and the instruction (README) to use the library within your project.
NEWS September 1, 2014: version 2.0 of the library is available. It can now be used with VisualStudio 2010 and Opencv 2.4.2 (or higher) versions.
|||L. Di Stefano, F. Tombari, S. Mattoccia, E. De Lisi, “Robust and accurate change detection under sudden illumination variations", ACCV'07 Workshop on Multi-dimensional and Multi-view Image Processing, 2007.|