Dynamic Learned Model Selection for Real-Time X-Ray Noise Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image processing techniques for X-ray movie imaging face challenges in real-time noise reduction due to difficulties in creating appropriate rules for various object structures, afterimage occurrence from noise reduction, and the need for multiple frames to be processed, which can lead to invalid exposure and delayed image processing.
Innovation Solution
An image processing apparatus that selects a learned model based on the number of obtained frames and performs inference processing using a multi-layer neural network, including a CNN, to apply noise reduction and other image processing techniques in real-time, even with only one frame available.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple frames are input to the neural network for noise reduction, then noise reduction performance is improved, but real-time processing capability deteriorates
Solution Approach 1:
The patent segments the learned models into multiple types based on the number of input frames they require. This allows the system to process frames in smaller batches rather than waiting for all frames to be collected, enabling real-time processing while still utilizing multi-frame noise reduction when sufficient frames are available.
Solution Approach 2:
The patent dynamically selects the appropriate learned model type based on the current frame count. When fewer frames are available, it selects models designed for smaller input sizes, and when more frames are available, it selects models optimized for larger input sizes. This dynamic adaptation resolves the contradiction between using multiple frames for better noise reduction and maintaining real-time processing capability.
2Measurement precision
If learned models requiring multiple frames are used, then noise reduction quality is improved, but processing delay increases
Solution Approach 1:
The patent prepares multiple learned models in advance, each trained to handle different numbers of input frames. This preliminary preparation allows the system to immediately select and apply the appropriate model based on the current frame count, eliminating the need to wait for additional frames to arrive before processing can begin.
3Device complexity
If a single learned model is used for all frame counts, then device complexity is reduced, but image processing quality deteriorates
Solution Approach 1:
The patent creates a universal framework that can handle variable frame counts by implementing a selection mechanism that chooses from multiple learned models. This multi-functional approach allows the system to adapt to different imaging scenarios and frame availability conditions while maintaining optimal image processing quality for each scenario.
Data Source
AI summary
An image processing apparatus configured to apply image processing to a moving image including a plurality of frames of radiation images is provided that includes: a selecting unit configured to select a learned model used for the image processing of a frame to be processed from among a plurality of learned models which differ in the number of frames to be input, based on the number of frames which have been obtained; and an inference processing unit configured to perform inference processing using the selected learned model in the image processing of the frame to be processed.