A method is provided for removing an illumination generated shadow in a captured image. An image is captured by an image capture device. Each pixel of the captured image is represented by a respective color value in a logarithmic graph. A non-linear illumination-invariant kernel is determined. An illumination direction for each respective color set is determined in the logarithmic graph that is orthogonal to the non-linear illumination-invariant kernel. A log-chromaticity value of each plotted pixel is projected on the non-linear illumination-invariant kernel. Edges are identified in the input image. Edges are identified in the illumination-invariant image domain. The identified edges are compared. A determination is made whether a shadow is present in response to an edge identified in the input image and an absence of a correlating edge in the illumination-invariant image domain. A shadow-reduced image is generated for scene analysis by a vehicle vision-based system.
展开▼
