Anomaly Exclusion Processing for Image Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing systems often detect false anomalies, leading to unnecessary computation costs due to overdetection of anomalies in printed matter, where an image not being an anomaly is mistakenly identified, and subsequent processes are performed even for these false detections.
Innovation Solution
An image processing apparatus and method that includes an anomaly detecting unit and an anomaly exclusion processing unit, which exclude anomalies based on specific conditions such as overlapping detection areas and differing types of anomalies, using area ratios and anomaly levels to determine which anomaly to exclude, thereby reducing false positives and computation costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If anomaly detection sensitivity is increased to detect more anomalies, then detection coverage is improved, but false anomaly detection increases leading to unnecessary computation
Solution Approach 1:
The patent applies preliminary action by performing anomaly exclusion processing before detailed anomaly analysis. The system pre-identifies and excludes false anomalies based on detection area overlap and type consistency, preventing unnecessary computation in subsequent processing steps. This early filtering mechanism resolves the contradiction by maintaining high detection sensitivity while eliminating false positives that would waste computational resources.
Solution Approach 2:
The patent introduces an intermediary mechanism - the anomaly exclusion processing unit - that acts as a mediator between anomaly detection and detailed anomaly analysis. This intermediary filters out false anomalies by checking detection area overlap ratios and type consistency, allowing the system to maintain high detection sensitivity without the computational overhead of analyzing false positives in detail.
2Reliability
If post processing is performed on all detected anomalies, then comprehensive analysis is achieved, but computation time increases due to false anomalies
Solution Approach 1:
The system performs preliminary anomaly exclusion before detailed post-processing by checking whether detection areas overlap and whether anomaly types are consistent. This preliminary filtering removes false anomalies from the processing queue, ensuring that comprehensive analysis is applied only to genuine anomalies, thereby maintaining reliability while reducing processing time.
Solution Approach 2:
The patent applies local quality by differentiating processing paths based on anomaly characteristics. Genuine anomalies undergo comprehensive post-processing analysis, while false anomalies are quickly identified and excluded through local checks on detection area overlap and type consistency. This selective processing maintains analysis completeness for true anomalies while minimizing time spent on false positives.
3Measurement precision
If anomaly detection threshold is lowered to reduce false negatives, then detection accuracy improves, but false positives increase requiring additional filtering
Solution Approach 1:
The anomaly exclusion processing unit serves as an intermediary that simplifies the overall system complexity. By introducing straightforward filtering criteria - detection area overlap ratio and anomaly type consistency - the system can use lower detection thresholds to improve anomaly detection accuracy without being overwhelmed by false positives. The intermediary handles the filtering burden, keeping the overall processing complexity manageable.
Solution Approach 2:
The patent changes key parameters for filtering - using detection area overlap ratio and anomaly type consistency as exclusion criteria. These parameter changes enable the system to maintain high detection sensitivity with lower thresholds while effectively filtering false positives through quantitative measures, thereby improving detection accuracy without proportionally increasing processing complexity.
Data Source
AI summary
An image processing apparatus includes an anomaly detecting unit configured to detect anomalies included in a target image; and an anomaly exclusion processing unit configured to exclude a specific anomaly among the detected anomalies. Further, the anomaly exclusion processing unit excludes one of an anomaly and another anomaly among the detected anomalies, if (a) a detection area of the anomaly, a detection area of the other anomaly, and an overlapping area of the detection areas of the anomaly and the other anomaly satisfy a predetermined condition and (b) a type of the anomaly and a type of the other anomaly are different from each other.