MRI Image Compression Using Deep Convolutional Adversarial Networks

Image compression is a type of data compression applied to digital images to reduce their cost for storage or transmission. Supervised learning with convolutional networks (CNNs) has seen widespread adoption in computer vision applications. Comparatively, unsupervised learning with CNNs has received less attention. In this research work, a class of CNNs called deep convolutional generative adversarial networks (DCGANs) has been developed. These networks have certain architectural constraints and demonstrate that they are strong candidates for unsupervised learning. Training on various image datasets shows convincing evidence that the proposed deep convolutional adversarial pair learns a hierarchy of representations, from object parts to scenes, in both the generator and discriminator. Additionally, this research uses the learned features for novel tasks, demonstrating their applicability as general image representations.