A computational approach to determination of main subject regions in photographic images

Which is the best way to organize/classify images by content?

“…The efficacy of this framework is demonstrated via three applications involving semantic understanding of pictorial images: (i) detection of the main photographic subjects in an image [22], (ii) selecting the most appealing image in an event, and (iii) classifying images into indoor or outdoor scenes. This last application refers specifically to the problem of scene classification [23].…”

Environmental consequence analyses of fish farm emissions related to different scales and exemplified by data from the Baltic – a review

“…If the boundary lines towards adjacent coastal areas and/or the open sea are drawn arbitrarily, one must also expect arbitrary modelling results. Various techniques, such as image segmentation, delineation and map regionalization for identifying natural units in spatial datasets have been used in the rapidly developing field of image processing (e.g., Cheng & Li, 2003;Kang, Engelke, & Kalender, 2004;Luo, Singhal, Etz, & Gray, 2004;van der Sande, de Jong, & de Roo, 2003) but applications for coastal delineation have to our knowledge yet not been developed. Segmentation is of interest in a variety of contexts (e.g., medicine, photography, remote sensing, GIS) where computational compartments are needed for descriptive or analytical purposes.…”

Main subject detection via adaptive feature selection

“…Researchers have proposed some methods to detect the main subject in an image. For example, the method of Luo et al [3] employs segmentation, perceptual clustering, and then feature extraction; the features are nally combined via a Bayesian network. Ma et al [4] perform MSD based on local contrast analysis followed by fuzzy growing.…”