2021/09/05 by Khawar Islam, Islam, Khawar, L. Minh Dang +6 · 1 citation
Computer Science · #Advanced Data Compression Techniques #Advanced Image Processing Techniques #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #FOS: Electrical engineering #Image and Signal Denoising Methods #Image and Video Processing (eess.IV) #Multimedia (cs.MM) #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.2109.01999
openalex publication_date 2021/09/05 · openalex created_date 2022/10/01 · openalex updated_date 2026/07/28
Image compression is a method to remove spatial redundancy between adjacent\npixels and reconstruct a high-quality image. In the past few years, deep\nlearning has gained huge attention from the research community and produced\npromising image reconstruction results. Therefore, recent methods focused on\ndeveloping deeper and more complex networks, which significantly increased\nnetwork complexity. In this paper, two effective novel blocks are developed:\nanalysis and synthesis block that employs the convolution layer and Generalized\nDivisive Normalization (GDN) in the variable-rate encoder and decoder side. Our\nnetwork utilizes a pixel RNN approach for quantization. Furthermore, to improve\nthe whole network, we encode a residual image using LSTM cells to reduce\nunnecessary information. Experimental results demonstrated that the proposed\nvariable-rate framework with novel blocks outperforms existing methods and\nstandard image codecs, such as George's ~ cite002 and JPEG in terms of image\nsimilarity. The project page along with code and models are available at\nhttps://khawar512.github.io/cvpr/\n