Autofocus Using Object Size Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Contrast Detection Auto Focus (CDAF) is limited by its slow speed due to the time required for determining lens movement direction and performing coarse searches, which is not acceptable for modern mobile phones and digital cameras, and existing faster methods like Phase Detection Auto Focus (PDAF) and Time of Flight (ToF) are costly and have reliability issues.
Innovation Solution
A method and apparatus for autofocusing that uses object size recognition based on pre-stored ground truth sizes, combined with pattern recognition technologies like face recognition, to calculate the distance between the lens and object, and adjust the lens position, thereby reducing the need for costly hardware and improving speed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If Contrast Detection Auto Focus (CDAF) is used, then the system is simple and cost-efficient, but the autofocus speed is slow
Solution Approach 1:
The patent applies preliminary action by using object size recognition to pre-determine the distance to the object before performing autofocus. The system recognizes objects in the scene, determines their sizes based on pre-stored ground truth data, calculates distances, and pre-positions the lens accordingly. This eliminates the need for slow iterative contrast detection, achieving 70-80% reduction in autofocus time while maintaining system simplicity and cost-efficiency without requiring expensive PDAF or ToF hardware
2Speed
If Phase Detection Auto Focus (PDAF) or Time of Flight (ToF) is used, then the autofocus speed is improved, but the hardware cost increases
Solution Approach 1:
The patent replaces the mechanical/optical hardware systems (PDAF sensors, ToF lasers) with a computational approach using object size recognition and pattern matching. Instead of using specialized hardware to determine distance, the system uses standard image sensors combined with software-based object recognition that compares detected object sizes against pre-stored ground truth databases to calculate distances and control focus, achieving comparable speed to PDAF/ToF without the associated hardware costs
Solution Approach 2:
The patent uses copying by creating virtual models of real objects with known sizes (ground truth databases) and comparing them against detected objects in the scene. The system copies the essential characteristic (size) of known objects to determine distance, eliminating the need for expensive specialized sensors while achieving accurate and fast autofocus through computational matching
3Measurement precision
If iterative coarse search is performed to determine lens movement direction, then the focus accuracy is improved, but the time consumption increases
Solution Approach 1:
The patent applies preliminary action by calculating the distance to the object in advance using object size recognition before the autofocus process begins. The system recognizes objects, determines their sizes, calculates distances using the formula distance = (object_size_in_scene / ground_truth_size) × focal_length, and pre-determines the required lens position. This eliminates the need for iterative coarse search and direction determination, achieving 70-80% reduction in autofocus time while maintaining accurate focus through pre-calculated lens positioning
Data Source
Figure 1A~2
Figure 3~4
Figure 5~6
AI summary
An autofocus method based on object size implemented in a digital image acquisition apparatus, comprising: acquiring a digital image of a scene, wherein the scene includes one or more objects; recognizing one or more objects based on pattern recognition; determining one or more sizes of the recognized object(s) based on a pre-stored database; calculating a distance to the recognized object(s) based on the determined size of the recognized object(s); adjusting a lens position based on the calculated distance.