A novel framework for image-to-image translation and image compression

Data-driven paradigms using machine learning are becoming ubiquitous in image processing and communications. In particular, image-to-image (I2I) translation is a generic widely used approach to problems, such as synthesis, style transfer, restoration. At the same time, neural compression has emerged data-driven alternative traditional coding approaches visual this paper, we study combination of these two into joint I2I framework, focusing on multi-domain synthesis. We first propose distributed by integrating quantization entropy an framework (i.e. I2Icodec). practice, functionality autoencoding) also desirable, requiring deploy alongside I2Icodec regular codec. Thus, further unified that allows both autoencoding capabilities single Adaptive residual blocks conditioned translation/compression mode provide flexible adaptation desired functionality. The experiments show promising results model.