Radiation Image Display Control for Processing Time Trade-off

The patent segments the image processing workflow into two distinct parts: (1) a fast preprocessing stage using traditional methods to quickly reduce noise and improve basic image quality, and (2) a subsequent machine learning-based processing stage that further refines image quality. This segmentation allows the system to achieve both quick initial results and high-quality final images, resolving the contradiction between processing speed and image quality.

The patent applies preliminary action by performing fast traditional image processing (such as filtering and noise reduction) before applying the more time-consuming machine learning algorithms. This preliminary processing prepares the image data in advance, reducing the computational burden on subsequent machine learning stages and enabling faster overall processing while maintaining high image quality.

The patent divides the noise reduction process into multiple stages with different algorithms optimized for different purposes. Traditional filtering methods are applied first for rapid noise suppression, followed by machine learning-based refinement. This segmentation enables the system to deliver noise-reduced images quickly while maintaining effectiveness, as each stage contributes differently to the overall noise reduction goal.

The patent ensures continuous useful action by implementing a pipeline where fast traditional processing and slower machine learning processing operate in sequence without idle time. The output of the first stage becomes the input of the second stage continuously, maximizing the utilization of processing resources and minimizing the total time from imaging to final display while maintaining effective noise reduction throughout the process.