Low-Light Image Enhancement via a Deep Hybrid Network

Area of Science:

• Computer Vision
• Image Processing
• Deep Learning

Background:

• Low-light conditions significantly degrade camera sensor image quality.
• Existing low-light image enhancement methods struggle with preserving structural details.

Purpose of the Study:

• To propose a novel trainable hybrid network for enhancing visibility in low-light images.
• To simultaneously learn global content and salient structures for improved image quality.

Main Methods:

• A hybrid network with two distinct streams: a content stream (encoder-decoder) and an edge stream.
• The edge stream utilizes a spatially variant recurrent neural network (RNN) guided by an auto-encoder to capture fine details.
• A unified network architecture integrates both streams.

Main Results:

• The proposed network effectively enhances the visibility of degraded low-light images.
• Experimental results demonstrate superior performance compared to state-of-the-art low-light enhancement algorithms.
• The hybrid approach successfully preserves both global content and structural details.

Conclusions:

• The developed trainable hybrid network offers a robust solution for low-light image enhancement.
• The dual-stream architecture effectively addresses the limitations of previous methods by preserving image structures.
• This work advances the field of image enhancement for challenging lighting conditions.