Image Compression Using Neural Networks

Main Article Content

S. YAGNASREE , A. SUBRAMANYAM , M. ANAND

Abstract

 


A large fraction of Internet traffic is now driven by requests from mobile devices with relatively small screens and often stringent bandwidth requirements. Due to these factors, it has become the norm for modern graphics-heavy websites to transmit low-resolution, low-byte count image previews (thumbnails) as part of the initial page load process to improve apparent page responsiveness. Increasing thumbnail compression beyond the capabilities of existing codecs is therefore a current research focus, as any byte savings will significantly enhance the experience of mobile device users. Toward this end, a general framework is proposed for image compression and a novel architecture based on convolutional and de-convolutional LSTM neural networks. This paper presents a set of full-resolution lossy image compression methods based on neural networks. Each of the architectures described can provide variable compression rates during deployment without requiring retraining of the network: each network need only be trained once. Proposed work is compared to previous work, showing improvements of 4.3%–8.8% AUC (area under the rate-distortion curve), depending on the perceptual metric used.

Article Details

Section
Articles