2021/02/12 by Mohamed Abid, Abid, Mohamed Abderrahmen, Ihsen Hedhli +3 · 1 citation
Computer Science · Engineering · Biochemistry, Genetics and Molecular Biology · #Advanced Image Processing Techniques #Image Processing Techniques and Applications #Cell Image Analysis Techniques
paper · pdf · doi:10.48550/arxiv.2102.06624
Traditionally, the main focus of image super-resolution techniques is on\nrecovering the most likely high-quality images from low-quality images, using a\none-to-one low- to high-resolution mapping. Proceeding that way, we ignore the\nfact that there are generally many valid versions of high-resolution images\nthat map to a given low-resolution image. We are tackling in this work the\nproblem of obtaining different high-resolution versions from the same\nlow-resolution image using Generative Adversarial Models. Our learning approach\nmakes use of high frequencies available in the training high-resolution images\nfor preserving and exploring in an unsupervised manner the structural\ninformation available within these images. Experimental results on the CelebA\ndataset confirm the effectiveness of the proposed method, which allows the\ngeneration of both realistic and diverse high-resolution images from\nlow-resolution images.\n