Autofocus Slope Correction via Adaptive Model Fitting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing autofocus systems are slow and often fail to converge on a proper focus, especially in low-contrast and low-light conditions, and are prone to errors due to distortion from saturation, cut-off, and noise.
Innovation Solution
An adaptive model fitting algorithm is used to correct the slope of the blur matching curve, combining multiple image estimations through a weighted mean approach to improve focusing accuracy and robustness, even in noisy conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional autofocus mechanisms search for a peak in the autofocus curve using image gradients, then the system can identify focus positions, but the process becomes slow and fails to converge properly in low-contrast and low-light conditions
Solution Approach 1:
The system performs preliminary calibration by capturing images at multiple known focus positions and pre-computing the relationship between focus position and image quality metric. This pre-computed model is then used during actual autofocus operations to directly determine focus positions without iterative searching, significantly reducing focusing time while maintaining accuracy.
Solution Approach 2:
The patent replaces the traditional mechanical iterative searching method with a computational approach using pre-computed blur matching curves and weighted mean calculations. Instead of physically moving the lens back and forth to search for focus peaks, the system uses image processing and mathematical modeling to directly calculate the optimal focus position.
2Reliability
If autofocus algorithms continuously adjust focus positions to search for the peak, then they may find the correct focus, but the system exhibits hunting behavior and fails to converge in challenging conditions
Solution Approach 1:
The system implements feedback by continuously monitoring the weighted mean focus position and the spread value, and adjusting the search range and step size based on these metrics. When the spread value indicates poor convergence, the algorithm automatically adjusts its parameters to improve stability, creating a self-regulating control system that prevents hunting behavior.
Solution Approach 2:
The autofocus algorithm dynamically adjusts its search parameters including step size, search range, and weighting factors based on real-time image quality metrics and convergence status. This dynamic adaptation allows the system to converge reliably across varying scene conditions without requiring overly complex fixed algorithms.
3Ease of manufacture
If the autofocus system uses simple peak detection methods, then the implementation remains simple, but the system is prone to errors from saturation, cut-off, and noise
Solution Approach 1:
The patent introduces an intermediary weighted mean calculation layer between the raw image data and the final focus decision. Instead of directly detecting peaks from potentially noisy image gradients, the system computes a weighted average of focus positions based on pre-computed blur matching curves, which filters out noise and saturation effects while maintaining implementation simplicity.
4Measurement precision
If multiple image estimations are combined without weighting, then the system processes more data for accuracy, but the computational load increases unnecessarily
Solution Approach 1:
The system applies local quality by using weighted averaging where recent or more reliable measurements are given higher weights. This allows the system to focus computational resources on the most informative data points rather than treating all measurements equally, achieving high accuracy with reduced computational load by selectively processing only the most valuable information.
Data Source
Figure 1~2B
Figure 3
Figure 4
AI summary
Autofocusing is performed in response to a weighted sum of previous blur difference depth estimates after being adaptively fitted at each focus adjustment iteration. Variance is also determined across both past and present estimations providing a confidence measure on the present focus position for the given picture. In one embodiment focus adjustment are repeated until the variance is sufficiently low as to indicate confidence that a proper focus has been attained. The method increases accuracy and speed of focusing by utilizing previous depth estimates while adapting the matching data to overcome distortion, such as due to saturation, cut-off and noise.