Seam carving is currently one of the most used methods to resize images. Its popularity is given by being a simple algorithm that maintains the content of the image. However, this technique is also used to remove objects and people from images, which can be a problem for Computational Forensics. The present proposal of scientific initiation aims at the creation of a seam carving process detection technique without having the knowledge of the original image, based on the image characteristics and image texture of properties provided by the Local Binary Pattern and using the Convolutional Neural Network techniques, which determines whether the image is original or changed by carving seam process. The proposed method of detection aims to serve as an aid to researchers in the field of computer forensics and cognitive computing, comparing its effectiveness with that of existing methods.
News published in Agência FAPESP Newsletter about the scholarship: