Adaptive Blurring Filter for Image Noise Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Small digital cameras, especially those with low-cost and portable devices, face challenges in noise reduction due to their limited image sensor capabilities, which affects image quality, especially in low-light conditions and requires processor-intensive methods that are not efficient for real-time noise identification and removal.
Innovation Solution
An intelligent blurring filter that uses pixel information from a single frame to reduce noise, employing a non-linear function and median filters to determine per-pixel blurring radii, balancing noise reduction with the preservation of fine image details, and operates without the need for multiple frames or significant processing resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If accurate noise determination methods are used, then noise identification accuracy is improved, but processing speed deteriorates due to processor-intensive operations
Solution Approach 1:
The patent employs a disposable, single-frame de-noising approach that processes each image independently without requiring multiple frames or complex temporal analysis. This single-use processing method reduces computational overhead while maintaining effective noise removal through localized gradient analysis and adaptive blurring.
Solution Approach 2:
The invention applies local quality by computing gradient magnitudes and applying adaptive blurring radii on a per-pixel basis rather than uniformly across the entire image. This localized processing allows accurate noise determination in each region while reducing overall computational complexity through parallelizable operations.
2Object-generated harmful factors
If blurring is applied to remove noise, then noise reduction is improved, but image detail preservation deteriorates
Solution Approach 1:
The patent implements dynamic adaptive blurring where the blur radius varies per pixel based on local gradient magnitude. Pixels with high gradients (edges and details) receive minimal or no blurring, while pixels with low gradients (smooth regions with noise) receive stronger blurring. This dynamic adaptation preserves image details while effectively reducing noise.
Solution Approach 2:
Different blurring strengths are applied to different regions of the image based on local characteristics. The gradient magnitude computation identifies regions requiring preservation (edges) versus regions suitable for noise reduction (smooth areas), enabling selective processing that maintains overall image quality.
3Measurement precision
If multiple frames are used for noise identification, then noise determination accuracy is improved, but processing complexity and resource requirements deteriorate
Solution Approach 1:
The invention uses a single-frame disposable approach where each image is processed independently without requiring frame buffering or temporal comparison. This eliminates the complexity of multi-frame processing pipelines while maintaining practical noise removal effectiveness through spatial gradient analysis.
Solution Approach 2:
The de-noising algorithm is self-sufficient, using only the information present in the single input frame to determine and remove noise. It does not require external reference frames or complex temporal analysis, making the processing simpler and more suitable for real-time applications on resource-constrained devices.
Data Source
AI summary
Image de-noising is described using an equalized gradient space. In one example, a method of de-noising an image includes determining an intensity gradient magnitude for an image, determining blurring radii for a plurality of pixels of the image using the intensity gradient, and blurring the image at each of the plurality of pixels using the blurring radii.