US Patent Application 18227120. Image Enhancement via Iterative Refinement based on Machine Learning Models simplified abstract
Contents
Image Enhancement via Iterative Refinement based on Machine Learning Models
Organization Name
Inventor(s)
Chitwan Saharia of Toronto (CA)
Jonathan Ho of Berkeley CA (US)
Mohammad Norouzi of Toronto (CA)
Image Enhancement via Iterative Refinement based on Machine Learning Models - A simplified explanation of the abstract
This abstract first appeared for US patent application 18227120 titled 'Image Enhancement via Iterative Refinement based on Machine Learning Models
Simplified Explanation
- The patent application describes a method for training a neural network to enhance images. - The method involves using a set of image pairs, where each pair consists of an image and a corresponding target version of the image. - The neural network is trained to predict an enhanced version of an input image based on the training data. - To train the neural network, a forward Gaussian diffusion process is applied to the target versions of the image pairs. - The forward Gaussian diffusion process adds Gaussian noise to the target versions, enabling iterative denoising of the input image. - The iterative denoising is based on a reverse Markov chain associated with the forward Gaussian diffusion process. - The trained neural network is then outputted as the result of the method.
Original Abstract Submitted
A method includes receiving, by a computing device, training data comprising a plurality of pairs of images, wherein each pair comprises an image and at least one corresponding target version of the image. The method also includes training a neural network based on the training data to predict an enhanced version of an input image, wherein the training of the neural network comprises applying a forward Gaussian diffusion process that adds Gaussian noise to the at least one corresponding target version of each of the plurality of pairs of images to enable iterative denoising of the input image, wherein the iterative denoising is based on a reverse Markov chain associated with the forward Gaussian diffusion process. The method additionally includes outputting the trained neural network.
