Comparison of Techniques for Mitigating the Effects of Illumination Variations on the Appearance of Human Targets
Several techniques have been proposed to date to build colour invariants between camera views with varying illumination conditions. In this paper, we propose to improve colour invariance by using data-dependent techniques. To this aim, we compare the effe
- PDF / 474,861 Bytes
- 12 Pages / 430 x 660 pts Page_size
- 36 Downloads / 185 Views
iversity of Technology, Sydney, Australia 2 ITC-CNR, Milano, Italy
Abstract. Several techniques have been proposed to date to build colour invariants between camera views with varying illumination conditions. In this paper, we propose to improve colour invariance by using data-dependent techniques. To this aim, we compare the effectiveness of histogram stretching, illumination filtration, full histogram equalisation and controlled histogram equalisation in a video surveillance domain. All such techniques have limited computational requirements and are therefore suitable for real time implementation. Controlled histogram equalisation is a modified histogram equalisation operating under the influence of a control parameter [1]. Our empirical comparison looks at the ability of these techniques to make the global colour appearance of single human targets more matchable under illumination changes, whilst still discriminating between different people. Tests are conducted on the appearance of individuals from two camera views with greatly differing illumination conditions and invariance is evaluated through a similarity measure based upon colour histograms. In general, our results indicate that these techniques improve colour invariance; amongst them, full and controlled equalisation consistently showed the best performance.
1
Introduction
Applications in the computer vision field that extract information about humans interacting with their environment are built upon the exploitation of appearance, shape and motion cues in videos. Appearance (i.e. colour-based) features are increasingly being used because cheaper, higher resolution cameras of good pixel quality are available. However, significant problems still affect the reliable use of appearance features for the analysis of humans in videos, such as the variations in illumination and the articulated nature of humans’ geometry. The goal of this paper is to improve the invariance of appearance features such as colour histograms for the global object. This is different from local colour invariants such as CSIFT [2] that describe the object’s colours only in a limited spatial neighbourhood. The improvement of colour invariance is investigated through the comparison of data-dependent techniques that compensate for illumination changes. The evaluation of the illumination invariance of these techniques is based upon measuring their ability to remain invariant for a single person under G. Bebis et al. (Eds.): ISVC 2007, Part II, LNCS 4842, pp. 116–127, 2007. c Springer-Verlag Berlin Heidelberg 2007
Comparison of Techniques for Mitigating the Effects of Illumination
50
20
50
50
50
100
60
150
100
100
80
150
200
100 150
200
250
140
250 20406080100
700
20
200
200 20406080100
500
600
400
150
500
40
20 40 60
600
1200
500
800
200
600
1500
100
200
0 400 0
200
0 400 0
500 100
200 200
0 400 0
1000
200
400
50
0
2000
300
100
20 40 60
400
100 300 200
20 40 60
1400
300
200
160
1000
400
150
120
150
300
0
50
40
100 100
117
200
Data Loading...
