Abstract
The present disclosure describes techniques to detect whether an image, e.g., a photograph, has been modified. For example, image modification can include application of one or more filters to the image. Further, the techniques described quantify the degree of change in the image due to the applied filters. One or more machine-learned models are trained and applied to determine the types of filters that are likely applied to an image. Training data for the models include unmodified images and the same images after one or more filters have been applied. The models are trained to determine the degree to which the image has been modified and to output probabilities that an image has been subjected to particular types of filters.
Recommended Citation
N/A, "Detecting modifications to images", Technical Disclosure Commons, (November 10, 2017)
https://www.tdcommons.org/dpubs_series/803