Low Complexity Change Detection via Dual Background Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing background estimation and subtraction methods for change detection in video processing are sensitive to noise and slow intensity variations, struggle to adapt to objects that have stopped or started moving, and incur high computational complexity due to performing motion detection on every input pixel.
Innovation Solution
The system employs two estimated background images with configurable block sizes and a confidence map to determine which background image to use for change detection, coupled with morphological filtering and connected component analysis to efficiently identify and classify changes, thereby reducing computational complexity while maintaining robustness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single estimated background image is used for change detection, then the system is simpler to implement, but it fails to adapt properly to objects that have stopped or started moving
Solution Approach 1:
The patent divides the single background estimation task into two separate background images: a first background image that adapts quickly to capture moving objects, and a second background image that adapts slowly to maintain stability. This segmentation allows the system to handle both moving and stationary objects effectively without excessive complexity.
Solution Approach 2:
The patent implements dynamic adaptability by allowing the system to switch between different background images based on motion detection. When motion is detected, the system uses the first background image for quick adaptation; when no motion is detected, it uses the second background image for stability. This dynamic switching resolves the contradiction between simplicity and adaptability.
2Measurement precision
If motion detection is performed on every input pixel, then detection precision is improved, but computational complexity increases significantly
Solution Approach 1:
The patent segments the image processing into two stages: first, block-level change detection that operates on larger regions to identify areas of interest, and second, pixel-level change detection that operates only on identified change regions. This segmentation maintains detection precision while significantly reducing computational complexity by avoiding full-pixel processing of the entire image.
Solution Approach 2:
The patent applies partial action by performing computationally intensive pixel-level change detection only on regions where changes have been detected at the block level, rather than processing every pixel in the entire image. This approach maintains precision in critical areas while reducing overall computational burden.
3Speed
If the background learns quickly from abandoned objects, then adaptation speed is improved, but noise sensitivity increases
Solution Approach 1:
The patent implements dynamic learning rates for the two background images: the first background image uses a higher learning rate (α1) for quick adaptation to moving objects, while the second background image uses a lower learning rate (α2) for stable, noise-resistant adaptation. This dynamic configuration allows the system to achieve both fast adaptation and noise immunity in different contexts.
Solution Approach 2:
The patent introduces a motion detection map as an intermediary that mediates between the two background images. This intermediary guides which background image should be used for updating based on detected motion, allowing the system to leverage the fast adaptation of the first background when motion is present and the noise resistance of the second background when motion is absent.
Data Source
AI summary
A method for low complexity change detection in a sequence of images using configurable block sizes is disclosed. In one embodiment, a first change detection map is generated by performing change detection based on configurable block sizes between a current image and one of an estimated first background image and a previous image. The first change detection map classifies each block as changed or unchanged. The selection between the previous image and the estimated background image for use in change detection is done using a confidence estimate, which is updated both at the low level and at the end of a high-level change analysis. In another embodiment, an estimated second background image is used in addition to the estimate first background image to help quickly adapt when a stationary object starts moving or when a scene object becomes stationary.