TY - JOUR
T1 - Image Retargetability
AU - Tang, Fan
AU - Dong, Weiming
AU - Meng, Yiping
AU - Ma, Chongyang
AU - Wu, Fuzhang
AU - Li, Xinrui
AU - Lee, Tong Yee
N1 - Funding Information:
To address the problems of the CAIR method selection and result evaluation, we introduce the notion of “image retargetabil-5,2019; acceptedJuly22,2019.DateofpublicationAugust1, 2019;dateManuscriptreceived August23,2018;revised January4, 2019and May ity” to quantitatively compute how well the image can be retar-of current version February 21, 2020. This work was supported in part by geted on the basis of its visual content. Fig. 1 shows the pre- National Key R&D Program of China under 2018YFC0807500, and in part dicted retargetability scores of several input images and the cor-61672520,and61702488,andin part byMinistryofScienceandTechnol-by NationalNaturalScienceFoundationofChinaunderGrants61832016, responding results of the “best” retargeting method selected by ogyunder108-2221-E-006-038-MY3,Taiwan,andinpartbyCASIA-Tencent our system.
Publisher Copyright:
© 1999-2012 IEEE.
PY - 2020/3
Y1 - 2020/3
N2 - Real-world applications could benefit from the ability to automatically retarget an image to different aspect ratios and resolutions while preserving its visually and semantically important content. However, not all images can be equally processed. This study introduces the notion of image retargetability to describe how well a particular image can be handled by content-aware image retargeting. We propose to learn a deep convolutional neural network to rank photo retargetability, in which the relative ranking of photo retargetability is directly modeled in the loss function. Our model incorporates the joint learning of meaningful photographic attributes and image content information, which can facilitate the regularization of the complicated retargetability rating problem. To train and analyze this model, we collect a dataset that contains retargetability scores and meaningful image attributes assigned by six expert raters. The experiments demonstrate that our unified model can generate retargetability rankings that are highly consistent with human labels. To further validate our model, we show the applications of image retargetability in retargeting method selection, retargeting method assessment and generating a photo collage.
AB - Real-world applications could benefit from the ability to automatically retarget an image to different aspect ratios and resolutions while preserving its visually and semantically important content. However, not all images can be equally processed. This study introduces the notion of image retargetability to describe how well a particular image can be handled by content-aware image retargeting. We propose to learn a deep convolutional neural network to rank photo retargetability, in which the relative ranking of photo retargetability is directly modeled in the loss function. Our model incorporates the joint learning of meaningful photographic attributes and image content information, which can facilitate the regularization of the complicated retargetability rating problem. To train and analyze this model, we collect a dataset that contains retargetability scores and meaningful image attributes assigned by six expert raters. The experiments demonstrate that our unified model can generate retargetability rankings that are highly consistent with human labels. To further validate our model, we show the applications of image retargetability in retargeting method selection, retargeting method assessment and generating a photo collage.
UR - http://www.scopus.com/inward/record.url?scp=85080866061&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85080866061&partnerID=8YFLogxK
U2 - 10.1109/TMM.2019.2932620
DO - 10.1109/TMM.2019.2932620
M3 - Article
AN - SCOPUS:85080866061
SN - 1520-9210
VL - 22
SP - 641
EP - 654
JO - IEEE Transactions on Multimedia
JF - IEEE Transactions on Multimedia
IS - 3
M1 - 8784158
ER -