Volume 10 Issue 1 - March 2018

  • 1. Surf descriptor based optimized face detection using boosted svm cascade and skin color filtering

    Authors : Jayasree M, N K Narayanan, Kabeer V, Arun C R

    Pages : 86-95

    DOI : http://dx.doi.org/10.21172/1.101.15

    Keywords : Face Detection, Boosted SVM Cascade, Skin color filtering, ROC Curve

    Abstract :

    An enhanced and optimized face detection system robust to variations in pose and illumination, along with minimum false positive rate has been proposed in this paper. SURF descriptors are used as feature vectors and SVM is used as the base classifier. A number of SVMs are trained using a boosting algorithm and stored as an ensemble of weighted SVMs. A cascade of different stages is built to make the final strong classifier. Optimization is a main criterion in designing this system, as far as false positive rate and processing time are concerned. These two features are achieved by a skin color filtering module integrated to the proposed face detection system, which outputs a skin-color segmented binary image. The current detection window is processed only if there are enough skin color pixels in that window. Further, the detection window is classified using boosted SVM cascade classifier. Typical post processing operations completes the face detection procedure. Testing on FDDB benchmark dataset shows that the proposed system outperforms classic boosted cascade face detection methods on ROC curve evaluation and state-of-the-art detectors on comparison based on detection rate.

    Citing this Journal Article :

    Jayasree M, N K Narayanan, Kabeer V, Arun C R, "Surf descriptor based optimized face detection using boosted svm cascade and skin color filtering ", https://www.ijltet.org/journal_details.php?id=928&j_id=4383, Volume 10 Issue 1 - March 2018, 86-95, #ijltetorg