Detecting semi-specular reflection component from a sequence of images

In this paper, I address the problem of detecting incidental reflection from a sequence of images to improve the quality of background subtraction. This paper begins with a brief overview of previous work on separating specular-reflection from the reflective surface. The model I propose is based on a dichromatic illumination model, while not making any assumptions about the reflectance spectra and illumination change. The model is mostly using pixel information and some context information of images such as the color of foreground object in the scene. Unlike the previous methods, I attempt to retrieve the specular-reflection coefficient in the chromaticity color space to differentiate the specular-reflection component from the foreground component. Following the explanation of the model, the results of the experiment can be found. I also offer some results from Linear Independent Component Analysis (ICA) and Blind Source Separation (BSS) for comparison’s sake. The distribution of the specularreflection coefficient shows that the linear discriminant function can be used in detecting the semi-specular reflection. This paper concludes with the significance of the model and poses the problems yet to be solved.