Paper Image

Recursive decomposition of images into illumination components

Published on:

2 April 2024

Primary Category:

Computer Vision and Pattern Recognition

Paper Authors:

Saurabh Saini,

P J Narayanan

Bullets

Key Details

Decomposes images into additive illumination components

Model-driven network with few parameters estimates components

Components enable targeted image enhancement regions

Improves state-of-the-art in low-light enhancement

Components beneficial for dehazing, deraining when combined with networks

AI generated summary

Recursive decomposition of images into illumination components

This paper presents a technique to decompose images into multiple illumination components in a recursive manner. A model-driven neural network is used to estimate these components, requiring only a small number of parameters. The components isolate regions with similar lighting characteristics, enabling targeted image enhancement. When applied to low-light image enhancement without paired training data, the method improves state-of-the-art performance. The illumination components also benefit other tasks like dehazing and deraining when combined with separate networks, demonstrating versatility.

Answers from this paper

Comments

No comments yet, be the first to start the conversation...

Sign up to comment on this paper

Sign Up