Jinxin Zhou

Title: Jinxin Zhou and Adaptive Super-Resolution

Introduction

Jinxin Zhou is an inventor associated with Columbus, Ohio, whose patent work addresses image super-resolution. His disclosed invention focuses on improving how diffusion models are trained for the task of generating higher-resolution images.

Latest Patents

Zhou is associated with 1 patent, titled “Adaptive Model for Super-Resolution.” The technology introduces a training framework intended to reduce the discrepancy between a diffusion model’s training and sampling processes.

One component, diffusion rectification, uses an estimated high-resolution image rather than a ground-truth high-resolution image as the seed for the forward process. This approach is designed to address performance issues arising when training and sampling rely on different data. The framework also uses estimation adaptation, injecting ground-truth information into noisy images to reduce training-estimation error. The amount of information injected varies according to each image’s position in the Markov chain.

Career Highlights

The available information identifies Zhou’s patent contribution in image processing and machine learning. His work applies diffusion-model techniques to the challenge of producing super-resolution images.

Collaborations

Zhou is associated with Microsoft Technology Licensing, LLC. Luming Liang and Tianyi Chen are identified as coworkers.

Conclusion

Jinxin Zhou’s “Adaptive Model for Super-Resolution” presents a diffusion-model training approach designed to improve image reconstruction by addressing training-sampling discrepancies and adapting ground-truth guidance.