Abstract
Given the diverse variety of device screens, approaches that rely on manual creation of various forms of an image for each display situation do not scale. Therefore, automated techniques are required to process and resize images to adjust for different devices on which the image is to be displayed. The goal of automated retargeting is to change the aspect ratio of the original image while preserving the semantic and visual meaning of the content within the image. Currently, typical retargeting processes involve a multitude of operations, use a variety of heuristics that are tuned manually, and do not work effectively for all images. This disclosure describes a generative adversarial network (GAN) used to retarget an input image to a different aspect ratio. The described approach involves scaling each pixel row within an image by a factor between 0 and 1.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Wang, Jingbin; Zabih, Ramin; Wang, Chen; Bowen, Richard Strong; and Wilber, Michael, "Automated image retargeting at scale using a generative adversarial network", Technical Disclosure Commons, (November 05, 2019)
https://www.tdcommons.org/dpubs_series/2649