How can I implement diffusion-based image-to-image translation in MATLAB, similar to the Brownian Bridge Diffusion Model (BBDM)?

MATLAB currently provides examples for Denoising Diffusion Probabilistic Models (DDPM) focused on image generation from noise.

However, I need to implement image-to-image translation, where the diffusion model transforms a source image into a target image (or a target domain).

In recent literature, the Brownian Bridge Diffusion Model (BBDM) (2022) has become a popular and effective choice for image-to-image tasks, because it builds a conditional Brownian bridge diffusion process between a source image and a target image.

I could not find any official MATLAB implementation related to :

• Brownian Bridge Diffusion Model
• diffusion-based image-to-image translation
• conditional diffusion with a source image input

Does MATLAB provide any built-in tools, functions, or examples for diffusion-based image-to-image translation?

(e.g., BBDM, DDIM inversion, Stable Diffusion–style image-to-image pipelines)

Has anyone implemented Brownian Bridge Diffusion or similar bridging-type models (e.g., Rectified Flow, DDIM inversion) in MATLAB?

Thanks!

[Ayush]

Hi Steven,

I could find the following documentation for image generation using diffusion:

https://www.mathworks.com/help/deeplearning/ug/generate-images-using-diffusion.html

However, I was unable to find a relevant MathWorks resource for BBDM. I would recommend you to contact MathWorks techincal support at:

https://www.mathworks.com/support/contact_us.html