Learning And Interaction Based Intrinsic Image Recovery

The observed color at any point on an object is influenced by many factors, including the shape and material of the object, the positions and colors of the light sources, and the position of the viewer. Two of the most important characteristics are the shading and reflectance of each point in the scene. The reflectance component represents how the material of the object reflects light independent of viewpoint and illumination, while the illumination accounts for shading effects, including shading due to geometry, shadows and interreflections. Intrinsic image algorithm aims at decompositing an image into reflectance image and shading image.Firstly, we propose an intrinsic image decomposition method through classification based on SVM. The shading and reflectance intrinsic images are found by classifying each derivative in the image as either being caused by shading ro a reflectance change. The derivative classifiers are found by training on a set of example shading and reflectance images based on DCT coefficient and color image information. Secondly, we propose a novel intrinsic image recovery approach using optimization. Our approach is based on the assumption of color characteristics in a local window in natural images. The decomposition is formulated by optimizing an energy function with adding a weighting constraint to every local image block. We get the intrinsic image result by resolving the optimization problem using Gaussian-Seidel iteration. The two methods were proved to be effective by the experiment results.