Foreground Detection Using Segmented Intensity and Chroma Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current foreground detection techniques in surveillance systems face challenges in achieving high accuracy while maintaining low operation loads, leading to increased hardware costs and complexity, especially when dealing with embedded platforms.
Innovation Solution
The proposed solution involves a foreground detection apparatus and method that uses a foreground characteristic value calculation module, a first filter, and a second filter to differentiate between foreground and background pixels based on image intensity and chroma similarity, reducing the operation load by pre-filtering out similar background and foreground elements before detailed filtering.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If color vector difference calculation is used for foreground detection, then detection accuracy is improved, but operation load is greatly increased
Solution Approach 1:
The patent divides the foreground detection process into two segments: a first filter that performs simple intensity difference comparison to eliminate obvious background pixels, and a second filter that performs more complex color vector difference calculation only on the remaining pixels. This segmentation allows the system to maintain high detection accuracy while significantly reducing the overall operation load by applying intensive calculations only where necessary.
Solution Approach 2:
The patent applies the partial action principle by using a two-stage filtering approach where the first filter handles the majority of pixels with simple intensity comparison, and only the necessary subset of pixels that pass through the first filter undergo the more computationally intensive color vector difference calculation in the second filter. This ensures that complex calculations are performed only partially, on exactly those pixels that require them for accurate foreground detection.
2Measurement precision
If high level processor is used to handle high operation load, then foreground detection accuracy is maintained, but product cost is increased
Solution Approach 1:
The patent segments the detection pipeline into two distinct filtering stages with different computational requirements. The first filter uses simple intensity thresholding that can be implemented with basic processing units, while the second filter applies color vector analysis only to a subset of pixels. This segmentation enables the use of lower-cost processors that can efficiently handle the distributed computational load without requiring high-end processing capabilities for the entire image.
Solution Approach 2:
By applying complex color vector calculations only to a partial set of pixels that pass through the first filter, the patent reduces the total number of operations that require high-performance processing. This partial application of intensive algorithms allows standard or low-cost processors to achieve the same detection accuracy that would otherwise require expensive high-level processors handling all pixels.
3Device complexity
If simple intensity difference method is used, then operation load is reduced, but detection accuracy deteriorates
Solution Approach 1:
The patent implements a segmented detection approach where the first filter uses simple intensity difference comparison to perform initial filtering, and the second filter applies color vector difference calculation to refine the detection results. This segmentation allows the system to combine the low operation load of simple methods with the high accuracy of complex methods, achieving both efficiency and precision in foreground detection.
Solution Approach 2:
The patent uses partial action by applying the simple intensity difference method to all pixels in the first filter, then applying the more accurate color vector difference method partially to only those pixels that require further analysis. This ensures that the detection system achieves high overall accuracy without bearing the full operation load of applying complex calculations to every pixel.
Data Source
AI summary
An apparatus and a method for fast foreground detection are provided. A foreground characteristic value calculation module calculates a foreground characteristic value by using an image intensity of each pixel in a target image and a background intensity of a corresponding pixel in a background model. A first filter determines a first threshold and a second threshold according to at least one scenic factor for capturing the target image and filters out non-foreground pixels having their foreground characteristic value between the first threshold and the second threshold from pixels in the target image. A second filter determines an image difference condition and a chroma similarity condition according to the scenic factor and filters out non-foreground pixels having their foreground characteristic value satisfying the image difference condition and the chroma similarity condition from pixels left by the first filter. Pixels left by the second filter are served as foreground pixels.