Object Detection Using Pixel-Luminance Differences
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing face detection methods are inefficient in processing large images due to the need for extensive computational operations and are limited in detecting objects of non-integer sizes, as they rely on integral images and rectangle features that only consider luminance differences between adjacent rectangular boxes.
Innovation Solution
A device and method using a plurality of weak discriminators that compute estimates based on the difference of luminance values of two pixels, allowing for high-speed object detection by employing a simple characteristic quantity and a weighted majority decision, enabling the detection of objects of any size and improving computational efficiency through sequential processing and suspension thresholds.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If integral images and rectangle features are used for face detection, then detection speed is improved, but the method is limited to detecting objects of integer sizes only
Solution Approach 1:
The patent changes the fundamental parameter from rectangular region-based features to pixel-pair difference features. By computing the difference between luminance values of two pixels at different positions, the method eliminates the constraint of integer-sized detection windows while maintaining high computational efficiency. This parameter change allows detection of objects of any size, not just integer multiples of a base size.
2Measurement precision
If multiple weak discriminators are used to improve detection accuracy, then discrimination capability is enhanced, but computational complexity increases
Solution Approach 1:
The patent segments the face detection problem into multiple independent pixel-pair comparisons. Each weak discriminator evaluates a specific pair of pixel positions, and the results are combined through weighted majority voting. This segmentation allows parallel computation of simple pixel differences, which is computationally cheaper than traditional rectangle feature methods, while achieving high accuracy through ensemble combination.
Solution Approach 2:
The patent employs a sequential processing strategy where weak discriminators are applied one by one until a confidence threshold is reached. This partial action approach avoids computing all possible pixel-pair differences, reducing computational complexity while maintaining high detection accuracy. The process stops early when sufficient confidence is obtained, making the system efficient in practice.
3Reliability
If traditional rectangle features are used, then detection robustness is improved, but processing time increases for large images
Solution Approach 1:
The patent extracts only the essential feature - the luminance difference between two specific pixels - rather than computing comprehensive rectangle features. By taking out only the critical pixel-pair difference needed for detection, the method achieves robust face detection with minimal computational overhead, significantly reducing processing time for large images while maintaining reliability.
Data Source
AI summary
An object detecting device for detecting an object in a given gradation image. A scaling section generates scaled images by scaling down a gradation image input from an image output section. A scanning section sequentially manipulates the scaled images and cutting out window images from them and a discriminator judges if each window image is an object or not. The discriminator includes a plurality of weak discriminators that are learned in a group by boosting and an adder for making a weighted majority decision from the outputs of the weak discriminators. Each of the weak discriminators outputs an estimate of the likelihood of a window image to be an object or not by using the difference of the luminance values between two pixels. The discriminator suspends the operation of computing estimates for a window image that is judged to be a non-object, using a threshold value that is learned in advance.