Image Processing Device Block-Based Noise Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image processing methods fail to accurately superimpose frames in live radiation images, leading to noise reduction issues due to misalignment and poor S/N ratio, resulting in incomplete noise elimination and image doubling.
Innovation Solution
An image processing device that divides the original and superimposition target images into target and superimposition target blocks, selects fusion blocks based on contrast and feature quantities, and connects these blocks to generate a noise reduction image, effectively aligning subject images and reducing noise.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If simple superimposition of continuous frames is performed, then noise reduction is achieved, but image doubling occurs due to subject movement and misalignment
Solution Approach 1:
The patent divides each frame into multiple blocks (e.g., 4 blocks) and processes them independently. This segmentation allows precise alignment of specific regions while maintaining overall noise reduction, preventing image doubling by ensuring each block is correctly positioned before superimposition.
Solution Approach 2:
The patent applies different processing strategies to different blocks based on their content and position. Each block is evaluated individually for alignment characteristics, and appropriate superimposition is performed only for blocks that meet the alignment criteria, ensuring high precision in critical regions while maintaining noise reduction benefits.
2Manufacturing precision
If pixel-by-pixel superimposition is performed to align images, then alignment precision is improved, but processing time increases and error recognition occurs
Solution Approach 1:
By dividing the image into blocks, the patent reduces the complexity of pixel-by-pixel comparison. Instead of processing the entire image pixel-by-pixel, only the blocks that need alignment are processed, significantly reducing processing time while maintaining precision through the block-based evaluation method.
Solution Approach 2:
The patent performs block-based alignment evaluation rather than complete pixel-by-pixel analysis. This partial action approach processes only the necessary portions of the image, reducing processing time while still achieving sufficient alignment precision for effective noise reduction.
3Reliability
If pixel values are changed boldly to remove strong noise, then noise elimination is improved, but image quality deteriorates due to over-processing
Solution Approach 1:
The patent evaluates each block individually to determine the appropriate superimposition strategy. Blocks with strong noise are identified and processed with more aggressive noise removal, while blocks with weaker noise or important structural information are processed more conservatively, preventing over-processing and maintaining image quality.
Solution Approach 2:
The patent uses contrast values and other image quality metrics as feedback to guide the noise removal process. The superimposition is adjusted based on the evaluated block characteristics, ensuring that noise elimination is applied selectively and not excessively, thereby maintaining overall image quality while removing strong noise components.
Data Source
AI summary
With the present invention, it is possible to provide an image processing apparatus capable of reliably removing noise, even for a live image having a poor S/N ratio. The present invention has a configuration to search for where, on a frame F0, a target block BT is reflected. When superimposing the frame F0 and a frame F1, the target block BT, which is a fragment of the frame F0, and a plurality of blocks to be superimposed, which is a fragment of the frame F1 are set, and if, from fusion blocks BF generated by superimposing the target block BT on each of the blocks to be superimposed BR, a selection block BS is selected wherein superimposed subject images most reinforce one another, it is possible to reliably suppress duplication of the subject images.