Foreground-background separation on GPU using order based approaches

Authors:
Raj Gupta;Reddy M. Sailaja;Swagatika Panda;Sushant Sharma;Anurag Mittal
Affiliations:
Indian Institute of Technology, Madras, Chennai, India;Indian Institute of Technology, Madras, Chennai, India;Indian Institute of Technology, Madras, Chennai, India;Indian Institute of Technology, Madras, Chennai, India;Indian Institute of Technology, Madras, Chennai, India
Venue:
Proceedings of the Seventh Indian Conference on Computer Vision, Graphics and Image Processing
Year:
2010

Citing 9
Cited 0

Non-parametric local transforms for computing visual correspondence

ECCV '94 Proceedings of the third European conference on Computer Vision (Vol. II)
Ordinal Measures for Image Correspondence

IEEE Transactions on Pattern Analysis and Machine Intelligence
Local Binary Patterns as an Image Preprocessing for Face Authentication

FGR '06 Proceedings of the 7th International Conference on Automatic Face and Gesture Recognition
An Intensity-augmented Ordinal Measure for Visual Correspondence

CVPR '06 Proceedings of the 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition - Volume 1
2D Euclidean distance transform algorithms: A comparative survey

ACM Computing Surveys (CSUR)
GPU Accelerated Multimodal Background Subtraction

DICTA '08 Proceedings of the 2008 Digital Image Computing: Techniques and Applications
SMD: A Locally Stable Monotonic Change Invariant Feature Descriptor

ECCV '08 Proceedings of the 10th European Conference on Computer Vision: Part II
A discriminative feature space for detecting and recognizing faces

CVPR'04 Proceedings of the 2004 IEEE computer society conference on Computer vision and pattern recognition
Block-Based methods for image retrieval using local binary patterns

SCIA'05 Proceedings of the 14th Scandinavian conference on Image Analysis

Quantified Score

Hi-index	0.00

Visualization

Abstract

Background modeling has been a challenging task in computer vision applications. Most of the approaches use the intensity space to do the background modeling. This basic assumption is not valid in the case of illumination changes. So, we have changed from the intensity space to the order space. We look on the patch (neighborhood of pixels) and build the model using the order among the pixels in it. The model built by us has more impact on the center part of the patch. Hence we have used the concept of overlapping of patches to make it more robust. The results have been shown on the standard PETS dataset as well as dataset collected by us using the camera setup in the outdoor environment. We have implemented it on GPU(NVIDIA Tesla C1060 Processor) to increase the throughput and we are able to achieve the 25X speed compared to CPU.