{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/learning-a-virtual-codec-based-on-deep","title":"Learning a Virtual Codec Based on Deep Convolutional Neural Network to Compress Image","arxiv_id":"1712.05969","date":"2017-12-16","proceeding":null,"authors":["Lijun Zhao","Huihui Bai","Anhong Wang","Yao Zhao"],"abstract":"Although deep convolutional neural network has been proved to efficiently\neliminate coding artifacts caused by the coarse quantization of traditional\ncodec, it's difficult to train any neural network in front of the encoder for\ngradient's back-propagation. In this paper, we propose an end-to-end image\ncompression framework based on convolutional neural network to resolve the\nproblem of non-differentiability of the quantization function in the standard\ncodec. First, the feature description neural network is used to get a valid\ndescription in the low-dimension space with respect to the ground-truth image\nso that the amount of image data is greatly reduced for storage or\ntransmission. After image's valid description, standard image codec such as\nJPEG is leveraged to further compress image, which leads to image's great\ndistortion and compression artifacts, especially blocking artifacts, detail\nmissing, blurring, and ringing artifacts. Then, we use a post-processing neural\nnetwork to remove these artifacts. Due to the challenge of directly learning a\nnon-linear function for a standard codec based on convolutional neural network,\nwe propose to learn a virtual codec neural network to approximate the\nprojection from the valid description image to the post-processed compressed\nimage, so that the gradient could be efficiently back-propagated from the\npost-processing neural network to the feature description neural network during\ntraining. Meanwhile, an advanced learning algorithm is proposed to train our\ndeep neural networks for compression. Obviously, the priority of the proposed\nmethod is compatible with standard existing codecs and our learning strategy\ncan be easily extended into these codecs based on convolutional neural network.\nExperimental results have demonstrated the advances of the proposed method as\ncompared to several state-of-the-art approaches, especially at very low\nbit-rate.","url_abs":"http://arxiv.org/abs/1712.05969v7","url_pdf":"http://arxiv.org/pdf/1712.05969v7.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"learning-a-virtual-codec-based-on-deep","repo_url":"https://github.com/mdcnn/mdcnn.github.io","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null}],"tasks":[{"task_slug":"blocking","task_name":"Blocking"},{"task_slug":"image-compression","task_name":"Image Compression"},{"task_slug":"quantization","task_name":"Quantization"},{"task_slug":null,"task_name":"valid"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1712.05969","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}