Illumination invariant segmentation of spatio-temporal images by spatio-temporal Markov random field model

In order to resolve occlusion problems, we have proposed the Spatio-Temporal Markov Random Field model for segmentation of spatio-temporal images. This S-T MRF optimizes the segmentation boundaries of occluded objects and their motion vectors simultaneously, by referring to textures and segment labeling correlations along the temporal axis as well as the spatial axes. Recently, we have extended this model to segment spatial MRF energy distributions of images along the temporal axis, in order to resolve the problem of illumination effects. Since such spatio-temporal distribution of MRF energies is so stable against illumination effects, the model is able to appropriately segment spatiotemporal images against variations in illumination and shading effects. Consequently, in the evaluation of S-T MRF for application to vehicle tracking in traffic images, segmentation boundaries of vehicle regions were successfully determined even in cases of serious occlusions and shading effects.