Photobomb Defusal Expert: Automatically Remove Distracting People From Photos

Ning Shan, Daniel Stanley Tan, Melkamu Sewuyie Denekew, Yung-Yao Chen, Wen-Huang, Kai-Lung Hua

Research output: Contribution to journalArticleAcademicpeer-review


Cropping is one of the main operations for removing unwanted or distracting elements in an image. It can portray the main subject in a better layout and enhance the image aesthetic for a better visual experience. However, manually cropping multiple images are tedious and time consuming. It also requires some amount of artistic skill to determine a good way to crop the image. In this paper, we propose an automatic photo cropping system that determines the optimal bounding box for cropping to produce aesthetically pleasing images. Our system also finds and removes distracting people to place the focus on the main subject. We combined both learned internal image representations using a convolutional autoencoder as well as manually extracted features to train our model. Experimental results of our system achieved significantly better performance compared to other existing automatic cropping methods.
Original languageEnglish
Pages (from-to)717-727
Number of pages11
JournalIEEE Transactions on Emerging Topics in Computational Intelligence
Issue number5
Publication statusPublished - Oct 2020
Externally publishedYes


Dive into the research topics of 'Photobomb Defusal Expert: Automatically Remove Distracting People From Photos'. Together they form a unique fingerprint.

Cite this