Image Selection and Combination for Noise Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing technologies face challenges in improving the signal-to-noise ratio (SNR) of digital images, especially in low light conditions, as noise reduction methods often compromise sharpness or are computationally expensive, and struggle with ghosting when combining images of moving objects.
Innovation Solution
A method that selects a subset of aligned sequential images and averages their pixel values to produce a combined image, leveraging face detection and saliency detection to prioritize sharpness in key regions while reducing noise, and allowing users to trade off noise reduction with sharpness through a single parameter.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If a low pass filter is run over an image to reduce noise, then noise is reduced, but textures in the scene are affected and sharpness is compromised
Solution Approach 1:
The image is divided into multiple sequential frames captured in burst mode. By segmenting the image processing task across multiple frames and selecting only aligned frames for combination, the patent reduces noise through averaging while preserving sharpness in aligned regions, avoiding the blurring effect of traditional low pass filters.
Solution Approach 2:
The patent applies noise reduction selectively by combining only aligned frames where corresponding regions match. This local approach ensures that noise is reduced in stationary regions while preserving sharpness in regions that maintain alignment, rather than uniformly applying filtering across the entire image.
2Object-affected harmful factors
If two or more images are combined to reduce noise, then noise reduction is achieved, but ghosting occurs due to misalignment
Solution Approach 1:
Before combining images, the patent performs preliminary alignment by detecting features (such as eyes or faces) across sequential frames and identifying which frames are properly aligned. This preliminary action ensures that only aligned frames are selected for combination, preventing ghosting artifacts while maintaining noise reduction benefits.
3Reliability
If portions of two or more images are combined to minimize ghosting, then ghosting is reduced, but computational complexity increases and noise reduction around moving objects is limited
Solution Approach 1:
The patent captures more frames than strictly necessary (up to 60 frames per second in burst mode) to ensure sufficient aligned frames are available for selection. By having excess frames to choose from, the system can select only the aligned portions without requiring complex frame-by-frame warping or blending operations, thus reducing computational complexity while maintaining effective noise reduction.
4Object-affected harmful factors
If multiple sequential images are processed to improve SNR, then signal to noise ratio is improved, but processing time and computational resources increase
Solution Approach 1:
The patent extracts and processes only the aligned frames from the burst sequence, excluding misaligned frames from the averaging operation. This selective extraction reduces the number of frames that require intensive processing, thereby improving signal-to-noise ratio through averaging while minimizing processing time by working with a subset of frames rather than all captured frames.
Data Source
AI summary
A method is disclosed that includes receiving multiple sequential images captured by an image capture device. The method includes selecting a subset of the multiple sequential images that are aligned to each other. The method further includes averaging pixel values from each image in the subset of the multiple sequential images to produce a combined image.