Aesthetics of photographs, photobooks, and magazine covers: Tools for autonomous quality evaluation and content creation
Abstract
Quantifying aesthetics of photographs has become a research topic of considerable interest. The first two chapters are dedicated to the investigation of aesthetic quality of general photographs as well as photos containing faces using machine learning approaches. Relevant features are constructed, and models are built for aesthetics quality class classification and aesthetics score regression. Having large number of personal photographs, being able to easily summarize and present them in an aesthetically pleasing form is desirable for users. In Chapter Three, we develop an automatic photobook creation algorithm that utilizes image content, metadata, composition, and predicted aesthetic scores to determine the selected set of images, and to suggest a proper layout for a photobook. Visual balance and visual weight are among the key considerations for graphic and document design. Among various factors that affect the visual weight of an element, we focus on the influence of color. In Chapter Four, psychophysical experiments are conducted to obtain from subjects the evaluation of visual weight relationships between pairs of sample colors. The numerical models of perceptual weights of colors are built, and they are utilized for the center-of-weight analysis of photographs and the style analysis of magazine covers in Chapter Five. In Chapter Six, a data-driven magazine cover design tool is developed by learning the design patterns from the dataset of professionally designed magazine covers. In particular, the probabilistic models for choosing masthead color and choosing the combination of byline colors given a background image are investigated.
Degree
Ph.D.
Advisors
Allebach, Purdue University.
Subject Area
Computer Engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.