ABSTRACT
This chapter presents a pixel-wise illuminant invariant image transform which is proposed based on the imaging model of cameras to deal with shadows, so that the transformed image is only related to the characteristics of cameras and captured objects ideally. The geometric reconstruction and road detection are processed interactively, for example the road region is segmented based on the reconstructed geometric information, and will direct the geometric reconstruction of the next row, improving the performance of matching accuracy and computational efficiency. The road region is reconstructed row-wisely from the bottom row to the row on which the road region disappears. Based on the road detection result of the current row, the region of interest is extracted so that the next row doesn't need to be processed in whole. In outdoor scenes, the lighting condition is complex and the effect on appearance caused by shadows influences the reliability of the road detection system.
