Machine learning on images using a string-distance
This work addresses image classification and clustering for satellite data, but it is incremental as it builds on an existing distance metric without demonstrating major breakthroughs.
The authors tackled the problem of image feature extraction by representing images as vectors of distances to prototype images using the Universal Image Distance, enabling automatic feature extraction without domain knowledge. They applied this method to satellite image data for both supervised and unsupervised learning tasks, though no concrete performance numbers were provided.
We present a new method for image feature-extraction which is based on representing an image by a finite-dimensional vector of distances that measure how different the image is from a set of image prototypes. We use the recently introduced Universal Image Distance (UID) \cite{RatsabyChesterIEEE2012} to compare the similarity between an image and a prototype image. The advantage in using the UID is the fact that no domain knowledge nor any image analysis need to be done. Each image is represented by a finite dimensional feature vector whose components are the UID values between the image and a finite set of image prototypes from each of the feature categories. The method is automatic since once the user selects the prototype images, the feature vectors are automatically calculated without the need to do any image analysis. The prototype images can be of different size, in particular, different than the image size. Based on a collection of such cases any supervised or unsupervised learning algorithm can be used to train and produce an image classifier or image cluster analysis. In this paper we present the image feature-extraction method and use it on several supervised and unsupervised learning experiments for satellite image data.