2016/11/14 by Subhajit Chaudhury, Chaudhury, Subhajit, Hiya Roy +1
Computer Science · Engineering · #Advanced Image Fusion Techniques #Advanced Image Processing Techniques #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Image Processing Techniques and Applications #Image and Signal Denoising Methods
paper · pdf · doi:10.48550/arxiv.1611.04481
openalex publication_date 2016/11/14 · openalex created_date 2022/10/04 · openalex updated_date 2026/07/28
We present a fully convolutional network(FCN) based approach for color image\nrestoration. FCNs have recently shown remarkable performance for high-level\nvision problem like semantic segmentation. In this paper, we investigate if FCN\nmodels can show promising performance for low-level problems like image\nrestoration as well. We propose a fully convolutional model, that learns a\ndirect end-to-end mapping between the corrupted images as input and the desired\nclean images as output. Our proposed method takes inspiration from domain\ntransformation techniques but presents a data-driven task specific approach\nwhere filters for novel basis projection, task dependent coefficient\nalterations, and image reconstruction are represented as convolutional\nnetworks. Experimental results show that our FCN model outperforms traditional\nsparse coding based methods and demonstrates competitive performance compared\nto the state-of-the-art methods for image denoising. We further show that our\nproposed model can solve the difficult problem of blind image inpainting and\ncan produce reconstructed images of impressive visual quality.\n