Transfer of aesthetic values in paintings to photographs
Date of Issue2014
School of Electrical and Electronic Engineering
Institute for Media Innovation
The current state-of-the-art research in computational vision and graphics offers many tools to change the look and feel of images. However, few academic studies are looking at how such operations can be deployed to improve the visual appeal of a photograph from an artistic viewpoint. This thesis focuses on studying the underlying attributes of visual appeal with the aid of pre-existent paintings, and developing algorithms and methods that facilitate the transfer of the organization of image attributes contributing towards aesthetics in paintings to photographs. One of the aesthetic considerations in paintings is the way the tone and color values have been organized with respect to various notions of contrast. In this thesis, the following contrasts are reviewed and discussed: global contrast for the creation of global mood, local contrast to highlight the local details, contrast across depth planes to enhance depth perception, regional contrast and center-corner contrast for the focusing of attention. The thesis investigates these contrasts in landscape and portrait paintings. Composition, which is subject to geometric attributes such as target positioning and spacing, is also an important aesthetic aspect of painting. The thesis presents algorithms and methods that improve the depth perception through transfer of contrast across depth planes based on the principle of atmospheric perspective effect, and improve the focus of attention through figure-ground regional contrast transfer and painter-style vignetting transfer. The thesis also covers the cropping of a portrait photograph to improve its visual appeal. Example-based approach is used in the thesis to handle the diversity of content in paintings and photographs. One of the contributions in this thesis addresses the enhancement of the atmospheric perspective effect in landscape photographs by the manipulation of depthaware lightness and saturation contrast values. Based on a study which shows that saturation contrast and lightness contrast inter- and intra- depth planes in paintings are purposefully organized and obey the principle of atmospheric perspective effect, an example-based method is developed to transfer such contrast organization in an example painting to a landscape photograph in an attempt to improve its visual appeal and illusion of depth. This contrast mapping is formulated as an optimization problem that simultaneously considers the desired inter-contrast, intra-contrast, and specific gradient constraints. Experimental results demonstrate the improved depth illusion and visual appeal in photographs. For portrait painting, the regional contrast is the main attribute that makes the figure, especially the face, stands out in the painting. As another contribution of this thesis, a method for the manipulation of the regional contrast in portrait photographs is developed which makes use of pre-modern portrait paintings as aesthetic examples. The contrast organization in the example painting is transferred to the photograph by mapping the inter- and intra- regional contrast values. A novel piecewise nonlinear transformation curve is proposed to achieve the contrast mapping. Experimental results demonstrate that by using this proposed method, the visual appeal of portrait photographs are effectively improved and the face and the figure become more salient. Vignetting is an effect that is manifested by being clear in the center and fading off at the edges by reducing the image’s brightness or saturation at the periphery. The thesis analyzes the vignetting effect in paintings and photographs. The observation of the difference shows that the vignetting effect in paintings is more purposely presented. An algorithm is then explored to apply the lightness weighting derived from the vignetting effect in an example painting to a photograph. The thesis also studies the composition of portrait paintings and develops an algorithm to improve the composition of a portrait photograph based on an example portrait painting. Pose, face direction and space around the target figure are the main elements considered in portrait composition. Space cropping technique is used to optimize the composition of the photograph based on the locations of body parts in the selected example painting.
DRNTU::Engineering::Electrical and electronic engineering::Electronic systems::Signal processing