Novel Methods for Extracting Illumination-invariant Images and Fast Classification of Textures

In this thesis we propose a standardized method for extracting illumination-invariant images and a novel approach for classifying textures. Experiments are also extended to include object classification using the proposed methods. The illumination-invariant image is a useful intrinsic feature latent in color image data. Existing methods of extracting the invariant image are dependent upon the characteristics of cameras. Here, assuming that every image consists of data in a standardized sRGB color space, we develop a standardized method for extracting the illumination-invariant that is independent of camera characteristics. Texture classification is an important aspect of Computer Vision. In this work, we greatly increase speed for texture classification while maintaining accuracy. Inspired by past work, we propose a new method for texture classification which is extremely fast due to the low dimensionality of our feature space. Finally, we classify images of objects captured by varying the illumination angle.