Image Processing Noise Reduction via Region-Specific Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing noise reduction methods in video signals are ineffective in dark parts due to increased noise levels with brightness, leading to improved noise reduction in bright areas but not in dark regions.
Innovation Solution
An image processing apparatus and method that selects a specific image-capturing mode to employ an applied noise model differing from a reference noise model, allowing for tailored noise reduction in predetermined signal-value regions like dark parts by transforming the noise model using a function that matches noise levels at boundaries.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If a reference noise model is employed for noise reduction in video signals, then noise reduction effect is improved in bright portions, but noise reduction is not effective in dark parts
Solution Approach 1:
The patent divides the image into different signal value regions (dark parts and other regions) and applies different noise models to each region. Specifically, a first noise model is used for dark parts while a second noise model (reference noise model) is used for other regions, allowing each region to receive optimized noise reduction treatment according to its characteristics.
Solution Approach 2:
The patent changes the noise model parameters based on signal value regions. By selecting different noise models (first noise model for dark parts, second noise model for other regions) according to the signal value, the system adapts the noise reduction parameters to match the local characteristics of each region, thereby improving overall noise reduction effectiveness.
2Ease of operation
If noise reduction is performed using a single noise model for all regions, then the processing is simple, but the noise reduction effect varies significantly between bright and dark regions
Solution Approach 1:
The patent implements region-specific noise modeling by categorizing pixels into dark parts and other regions based on signal values, then applying appropriate noise models to each category. This localized approach ensures that each region receives noise reduction processing optimized for its specific characteristics.
Solution Approach 2:
The patent segments the image processing into distinct pathways: one for dark parts using a first noise model and another for other regions using a second noise model. This segmentation allows independent optimization of noise reduction for each region type while maintaining overall system coherence.
Data Source
AI summary
An object is to improve noise reduction. Provided are an image-capturing-mode selecting portion (104) that selects one image-capturing mode from a plurality of image-capturing modes; and a noise-reduction processing portion (110) that performs noise-reduction processing for input image signals by employing, in the case in which the image-capturing mode selected by the image-capturing-mode selecting portion (104) is a specific image-capturing mode, an applied noise model in which a noise model employed for a predetermined signal-value region differs from a reference noise model determined on the basis of properties of an image-acquisition element.