Exploiting Neighbors for Faster Scanning Window Detection in Images - Advanced Concepts for Intelligent Vision Systems - page 214

Exploiting Neighbors for Faster Scanning
Window Detection in Images

Pavel Zemc ık, Michal Hradis, and Adam Herout

Graph@FIT, Brno University of Technology, Bozetechova 2, Brno, CZ
{ zemcik,ihradis,herout } @fit.vutbr.cz

Abstract. Detection of objects through scanning windows is widely
used and accepted method. The detectors traditionally do not make use
of information that is shared between neighboring image positions al-
though this fact means that the traditional solutions are not optimal.
Addressing this, we propose an ecient and computationally inexpen-
sive approach how to exploit the shared information and thus increase
speed of detection. The main idea is to predict responses of the classi-
fier in neighbor windows close to the ones already evaluated and skip
such positions where the prediction is confident enough. In order to pre-
dict the responses, the proposed algorithm builds a new classifier which
reuses the set of image features already exploited. The results show that
the proposed approach can reduce scanning time up to four times with
only minor increase of error rate. On the presented examples it is shown
that, it is possible to reach less than one feature computed on average
per single image position. The paper presents the algorithm itself and
also results of experiments on several data sets with different types of
image features.

1 Introduction

Scanning window technique is commonly used in object detection in images. In
combination with highly selective and fast classifiers, it provides state-of-the-
art success rates under real-time constraints for various classes of target objects
[14,6,3]. Although, in reality, much information is shared between neighboring
(overlapping) image positions, they are normally classified independently. Mak-
ing use of this shared information has a potential to reduce amount of compu-
tations during scanning.
In this paper, we propose an effective and at the same time simple and compu-
tationally inexpensive method which uses the dependency between neighboring
image position to suppress computing the original detection classifier at nearby
locations. The proposed method learns a new classifiers which predict the re-
sponses of the original detection classifier at neighboring positions. When the
prediction is confident enough, computing the original classifier is suppressed.
We propose to use WaldBoost algorithm [11] to learn the suppressing classi-
fiers in such way that they reuse computations of the original detection classifier.
These reused computations can be image features in case of Viola & Jones' [14]