Saliency Based Image Cropping

Abstract

Image cropping is a technique that is used to select the most relevant areas of an image, discarding the useless ones. Handmade selection, especially in case of large photo collections, is a time consuming task. Automatic image cropping techniques may help users, suggesting to them which part of the image is the most relevant, according to specific criteria. We suppose that the most visually salient areas of a photo are also the most relevant ones to the users. In this paper we present an extended version of our previously proposed method, to extract the saliency map of an image, which is based on the analysis of the distribution of the interest points of the image. Three different interest point extraction algorithms are evaluated within an automatic image cropping system, to study the effectiveness of the related saliency maps for this task. We furthermore compared our results with two state of the art saliency detection techniques. Tests have been conducted onto an online available dataset, made of 5000 images which have been manually labeled by 9 users.

Preference for Well-Balanced Saliency in Details Cropped from Photographs

“…Based on saliency calculations of low-level image properties, a large number of computational tools have been devised to carry out cropping decisions automatically (also called image retargeting ; for reviews, see Vaquero et al, 2010 ; Ardizzone et al, 2013 ). For example, Ardizzone et al ( 2013 ) compared five different saliency algorithms for automatic cropping and obtained favorable results with all of them.…”

“…Based on saliency calculations of low-level image properties, a large number of computational tools have been devised to carry out cropping decisions automatically (also called image retargeting ; for reviews, see Vaquero et al, 2010 ; Ardizzone et al, 2013 ). For example, Ardizzone et al ( 2013 ) compared five different saliency algorithms for automatic cropping and obtained favorable results with all of them. Two other studies on saliency-based cropping applications were published by Suh et al ( 2003 ) who shrank original images to produce easily recognized thumbnails for image retrieval, and by Ciocca et al ( 2007 ) who redesigned large images for small screens with an adaptive visual attention model that incorporated semantic information.…”

Exploring Salient Thumbnail Generation for Archival Collections Online

“…There is also some evidence that the Itti-Koch model offers real advantages for thumbnail generation besides its accessibility. Compared to other models evaluated for their use in automatic cropping, Itti-Koch scored particularly high on precision (Ardizzone et. al., 2013).…”

Image Aesthetic Assessment: An experimental survey

“…A majority of cropping schemes in the literature can be divided into three main approaches. Attention/Saliency-based approaches [101], [102], [103] typically extract the primary subject region in the scene of interest according to attention scores or saliency maps as the image crops. Aesthetics-based approaches [104], [105], [106] assess the attractiveness of some proposed candidate crop windows with low-level image features and rules of photographic composition.…”