Temporal De-noising via Persistent Image Buffer and Noise Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current camera systems face significant challenges in producing temporally stable images due to noise fluctuations, particularly shot noise and read noise, which affect image quality and accuracy, especially in low-light conditions and when small parts of the image or individual pixels are examined.
Innovation Solution
The implementation of a computational/statistical model that estimates noise by simulating the inverse behavior of the camera pipeline, incorporating noise models that account for variables like electron well capacity, gain, and exposure time, and uses a persistent image buffer to blend data from previous frames, thereby reducing noise and stabilizing pixel values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If temporal de-noising operations are applied to reduce noise in image frames, then image quality and temporal stability are improved, but computational cost increases
Solution Approach 1:
The patent pre-calculates and stores noise model parameters (electron well capacity, gain, exposure time) during camera pipeline operations. These pre-computed values are then reused during temporal de-noising operations, avoiding redundant calculations and reducing real-time computational burden while maintaining de-noising effectiveness
Solution Approach 2:
The patent introduces a persistent image buffer as an intermediary data structure that stores historical pixel values across multiple frames. This buffer serves as a mediator between current frame data and noise reduction algorithms, enabling efficient temporal comparisons and statistical noise estimation without requiring complex real-time processing of entire image histories
2Measurement precision
If noise models simulating camera pipeline inverse behavior are implemented, then measurement precision of noise estimation is improved, but device complexity increases
Solution Approach 1:
The patent divides the noise estimation process into separate modular components: electron well capacity calculation, gain adjustment, exposure time normalization, and temporal blending. Each component handles a specific aspect of noise modeling independently, making the complex noise estimation process more manageable and implementable in camera pipelines
Solution Approach 2:
The patent dynamically adjusts noise model parameters (electron well capacity, gain, exposure time) based on actual camera pipeline conditions and scene characteristics. By changing these parameters adaptively rather than using fixed values, the model achieves higher noise estimation accuracy across varying lighting conditions and camera settings without requiring a fundamentally more complex model structure
Data Source
AI summary
Methods, systems, devices and computer software/program code products enable generating reduced-noise image frames based on image frames received from a digital, camera pipeline; and enable efficient stereo image search between corresponding images generated by at least two cameras.