Detection of objects in an image using self similarities

Abstract

An image processor (10) has a window selector for choosing a detection window within the image, and a self similarity computation part (40) for determining self-similarity information for a group of the pixels in any part of the detection window, to represent an amount of self-similarity of that group to other groups in any other part of the detector window, and for repeating the determination for groups in all parts of the detection window, to generate a global self similarity descriptor for the detection window. A classifier (50) is used for classifying whether an object is present based on the global self-similarity descriptor. By using global self-similarity rather than local similarities more information is captured which can lead to better classification. In particular, it helps enable recognition of more distant self-similarities inherent in the object, and self-similarities present at any scale.

Description

FIELD OF THE INVENTION[0001]This invention relates to apparatus and methods for image processing to detect objects such as humans, and to corresponding computer programs for carrying out such methods and to memory devices storing the computer programs and also to corresponding integrated circuits.BACKGROUND OF THE INVENTION[0002]Pedestrian detection has been a focus of recent research due to its importance for practical applications such as automotive safety [see refs 11, 8] and visual surveillance [23]. The most successful model to date for “normal” pedestrians, who are usually standing or walking upright, is still a monolithic global descriptor for the entire search window. With such a model, there are three main steps which can be varied to gain performance: feature extraction, classification, and non-maxima suppression. The most common features extracted from the raw image data are variants of the HOG framework, i.e. local histograms of gradients and (relative) optic flow [3, 4,...

AI Technical Summary

Benefits of technology

The patent text explains a method called global self-similarity that can capture more information and improve classification. This method can recognize self-similarities in objects at a distance and at different scales. By using this method, the classifier can capture the most discriminant self-similarities and improve the accuracy of object recognition.

Problems solved by technology

These recent efforts to record data of realistic complexity have also shown that there is still a gap between what is possible with pedestrian detectors and what would be required for many applications: in [6] the detection rate of the best methods is still <60% for one false positive detection per image, even for fully visible people.

Images