An Unscented Kalman filter based novel face detector and its robust system for illumination variant images using stochastic resonance
| Autor: | Noriko Kojima, Bikash Lamsal, Naofumi Matsumoto |
|---|---|
| Rok vydání: | 2016 |
| Předmět: |
business.industry
Computer science Detector ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION Kalman filter White noise Image enhancement Film grain noise Haar-like features Computer Science::Computer Vision and Pattern Recognition Viola–Jones object detection framework Computer vision Artificial intelligence business Face detection |
| Zdroj: | ICCCNT |
| DOI: | 10.1145/2967878.2967887 |
| Popis: | This paper is associated with the two special approaches on face detection. Firstly, we develop a novel face detector for detecting the human faces in a still image by using the unscented Kalman filter (UKF). The first approach or the face detector is developed by combining the modified skin color detector, modified Viola Jones eye detector, modified Haar cascade classifier and the UKF. The UKF is used for removing the film grain noise from the image and increasing the face detection rate. The UKF is being used for the very first time in the face detector and thus reflects the novelty of our current research. Secondly, we develop a robust system of our face detector by using our "proposed stochastic resonance (SR) technique" for image enhancement and detecting the presence of face on the illumination variant images. Our "proposed SR technique" extracts the image candidates from the dark and illumination variant images and makes the image visible and clear by applying the Gaussian white noise in an image, keeping the threshold value at constant zero. The resultant images are then summed up in a series for extracting the image candidates from the illumination variant images and make them visible. The idea of our "proposed SR technique" has been developed from the summing network process of the stochastic resonance noise. The images, once enhanced are passed to our face detector for detecting the presence of the face in an image and clarify the effectiveness of our current research. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|