Color Image Compression Using a Learned Dictionary of Pairs of Orthonormal Bases

Summary form only given. We present efficient machine learning methods for color image compression which simultaneously learn bases for compact image representation as well as the color space. We show the benefits of representing color image patches as 2D matrices of size n × 3n rather than as 3D patches of size n × n × 3. We also present a method to leverage greater representational power from a learned dictionary without increasing its size.