StereoDiffusion: Training-Free Stereo Image Generation Using Latent Diffusion Models

TL;DR

StereoDiffusion offers a training-free, efficient method for generating high-quality stereo image pairs by modifying latent variables within the Stable Diffusion model, suitable for XR device applications.

Contribution

It introduces a novel, training-free approach that integrates seamlessly with Stable Diffusion to generate stereo images without fine-tuning or post-processing.

Findings

Achieves state-of-the-art quantitative scores in stereo image generation.

Enables fast, training-free stereo image pair creation.

Maintains high image quality throughout the process.

Abstract

The demand for stereo images increases as manufacturers launch more XR devices. To meet this demand, we introduce StereoDiffusion, a method that, unlike traditional inpainting pipelines, is trainning free, remarkably straightforward to use, and it seamlessly integrates into the original Stable Diffusion model. Our method modifies the latent variable to provide an end-to-end, lightweight capability for fast generation of stereo image pairs, without the need for fine-tuning model weights or any post-processing of images. Using the original input to generate a left image and estimate a disparity map for it, we generate the latent vector for the right image through Stereo Pixel Shift operations, complemented by Symmetric Pixel Shift Masking Denoise and Self-Attention Layers Modification methods to align the right-side image with the left-side image. Moreover, our proposed method maintains a high standard of image quality throughout the stereo generation process, achieving state-of-the-art scores in various quantitative evaluations.