BlendGAN: Learning and Blending the Internal Distributions of Single Images by Spatial Image-Identity Conditioning

TL;DR

BlendGAN is a novel generative framework that learns and blends internal distributions of multiple images using spatial conditioning, enabling advanced image editing and fusion tasks beyond single-image capabilities.

Contribution

It introduces a single model that learns multiple internal image distributions with spatial conditioning, facilitating image blending and morphing.

Findings

Enables morphing and melding of multiple images.

Supports structure-texture fusion between images.

Extends single-image generative models to multiple images.

Abstract

Training a generative model on a single image has drawn significant attention in recent years. Single image generative methods are designed to learn the internal patch distribution of a single natural image at multiple scales. These models can be used for drawing diverse samples that semantically resemble the training image, as well as for solving many image editing and restoration tasks that involve that particular image. Here, we introduce an extended framework, which allows to simultaneously learn the internal distributions of several images, by using a single model with spatially varying image-identity conditioning. Our BlendGAN opens the door to applications that are not supported by single-image models, including morphing, melding, and structure-texture fusion between two or more arbitrary images.