Intrinsic Image Extraction from a Single Image

Intrinsic image extraction has long been an important task for computer vision applications. However, this task is nontrivial at all because it is an ill-conditioned problem. The technique proposed in this paper uses one single image to achieve the purpose. The proposed approach first convolves the input image with a prescribed set of derivative filters. The pixels of the derivative images are next classified as being reflectance-related or illumination-related according to three measures: chromatic, intensity, and sharpness measures, calculated in advance for pixels from the input image. A de-convolution process is finally applied to the classified derivative images to obtain the intrinsic images of the original image. Both synthetic and real images have been utilized in our experiments. The results reveal the feasibility of the proposed technique in both rapidly and effectively decomposing intrinsic images from one single image.