Video Illumination Enhancement via Neural Network Segmentation

The patent segments the image processing task by dividing images into different illumination categories (low illumination, ultra-low illumination, normal illumination) and applying different neural network models to each category. This segmentation allows each neural network to be specialized for specific image types, improving processing efficiency and quality while maintaining manageable system complexity through modular architecture.

The patent implements a dynamic selection mechanism that automatically chooses the appropriate neural network model based on the illumination characteristics of each input image. The system dynamically categorizes images into different illumination groups and selects the corresponding specialized neural network, providing adaptability to various illumination conditions while maintaining ease of operation through automated classification and selection processes.

The patent applies partial action by selectively enhancing only those video frames that require illumination improvement. The system classifies each frame according to its illumination characteristics and applies neural network enhancement only to low illumination and ultra-low illumination frames, while normal illumination frames are processed more simply or skipped. This approach maintains overall video quality while significantly reducing processing time and energy consumption compared to enhancing all frames.