Enhancement and stylization of photographs

Abstract

A photograph captured by a digital camera may be the final product for many casual photographers. However, for professional photographers, this photograph is only the beginning: experts often spend hours on enhancing and stylizing their photographs. These enhancements range from basic exposure and contrast adjustments to dramatic alterations. It is these enhancements - along with composition and timing - that distinguish the work of professionals and casual photographers. The goal of this thesis is to narrow the gap between casual and professional photographers. We aim to empower casual users with methods for making their photographs look better. Professional photographers could also benefit from our findings: our enhancement methods produce a better starting point for professional processing. We propose and evaluate three different methods for image enhancement and stylization. First method is based on photographic intuition and is fully automatic. The second method relies on expert's input for training; after the training this method can be used to automatically predict expert adjustments for previously unseen photographs. The third method uses a grammar-based representation to sample the space of image filter and relies on user input to select novel and interesting filters.