CAVALCANTI, C. S. V. C.; http://lattes.cnpq.br/2531237708249588; CAVALCANTI, Cláudio Sebastião Vasconcelos da Cunha.
Resumo:
The popularization of digital photography, due to the continued decline in the costs
of photographic equipments, has resulted in a demand for tools and methods to facilitate
the selection and organization of very large collections of photographs. In such
a context, this thesis aims at conceiving methods for analysis, selection, clustering,
and layouting of photo albums. Aiming the photo selection, methods for photographic
composition analysis are proposed. In particular, i t is proposed a method for analyzing
the negative space of a photograph. I t is also investigated the analysis of the color
harmony of a photograph. Such analysis required an image segmentation algorithm,
which was developed by adopting an evolutionary computation strategy. Photo clustering
is achieved by the use of photo timestamps obtained from images metadata.
Additionally, the image color distribution is used in order to provide a content-based
photo clustering. For the layouting process of digital photo albums, i t is proposed a
method for content-aware cropping of digital photographs. Such method uses a set of
features (e.g., human skin-tone and detected face coordinates) for defining areas with
salient regions, which should be preserved. A subjective evaluation was performed in
order to evaluate the proposed methods. The experimental results show that the proposed
method is promising because they are correlated with human decisions. It was
also noted that the employed evolutionary method produced an image segmentation
algorithm which is adequate for its purposes, even using a small set of image processing
tools. Finally, a final experiment was performed, with the whole system integrated,
and it was shown that 8 1% of the participants agreed that it significantly reduces the
effort in producing a photo-book.