Foreground Detection Under Illumination Variation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image processing methods fail to accurately detect foreground objects under varying environment illumination, especially when visual elements are occluded, leading to false detection and reduced accuracy in foreground detection.

Innovation Solution

The method updates both matching and non-matching visual element models in the scene model immediately following changes in environment illumination, using a first updating unit for matching models and a second updating unit for non-matching models, ensuring the scene model adapts to illumination variations, thereby preventing false detection of occluded elements.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional background subtraction method is used to detect foreground, then detection speed is maintained, but false foreground detection occurs under illumination variation

Engineering Contradiction:
Improveforeground detection accuracyVSAvoidillumination variation influence
Core Design Contradiction:
Measurement precisionVSObject-affected harmful factors

Solution Approach 1:

The patent applies preliminary action by pre-establishing a scene model that includes multiple visual element models representing different appearance states (occluded, unoccluded, different illuminations). Before foreground detection, the system prepares these pre-trained models to quickly compare against current video frames, enabling rapid adaptation to illumination changes without complex real-time calculations.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent employs parameter changes by dynamically adjusting the scene model parameters based on detected visual elements. When a visual element is detected, the system updates the appearance features (color, texture, illumination parameters) of corresponding visual element models in the scene model, allowing the model to adapt to changing illumination conditions and reduce false foreground detection.

Inventive Principle:
Principle #35Parameter changes

2Adaptability or versatility

If visual element model is updated only when matching current frame, then model stability is maintained, but adaptation to illumination changes is delayed

Engineering Contradiction:
Improveillumination adaptation speedVSAvoidscene model stability
Core Design Contradiction:
Adaptability or versatilityVSStability of the object's composition

Solution Approach 1:

The patent applies segmentation by dividing the scene model into multiple independent visual element models, each representing different appearance states (occluded, unoccluded, different illuminations). This segmentation allows selective updating of specific visual element models based on detection results, maintaining overall model stability while enabling targeted adaptation to illumination changes.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent implements dynamics by making the scene model adaptive and dynamic rather than static. The system continuously updates visual element models in the scene model based on detected visual elements and their appearance features, allowing the model to dynamically adapt to changing illumination conditions while maintaining stability through structured update mechanisms.

Inventive Principle:
Principle #15Dynamics

3Measurement precision

If strict threshold comparison is used for foreground detection, then detection precision is maintained, but false foreground occurs when element reappears after occlusion

Engineering Contradiction:
Improveforeground detection precisionVSAvoiddetection reliability under occlusion
Core Design Contradiction:
Measurement precisionVSReliability

Solution Approach 1:

The patent applies copying by creating and maintaining multiple copies of visual element models representing different appearance states (occluded, unoccluded, different illuminations) in the scene model. When a visual element reappears after occlusion, the system can compare against the appropriate model copy that matches the current appearance state, reducing false foreground detection while maintaining detection precision.

Inventive Principle:
Principle #26Copying

Data Source

PatentEP3543902B1Image processing apparatus and method and storage medium storing instructions
Publication Date: 2021.06.02 CANON KK
  • EP3543902B1 patent drawingFigure 1
  • EP3543902B1 patent drawingFigure 2
  • EP3543902B1 patent drawingFigure 3

AI summary

The present invention discloses an image processing apparatus and method and storage medium storing instructions. The image processing apparatus comprises a unit for acquiring a current video frame and a scene model based on an input video; a unit for updating visual element models in said scene model, which match with the visual elements in said current video frame, based on the visual elements in said current video frame; and a unit for updating visual element models in said scene model, which do not match with the visual elements in said current video frame, based on the visual elements in said current video frame and in the previous video frames thereof. According to the present invention, the whole scene model can vary immediately following the variation in the environment illumination.