A Variational Approach To Pixel-Wise Adaptive Threshoding For An Accurate Moving Object Detection
In computer vision application, object detection is fundamental and is an important step for video analysis.
Several works aimed at detecting objects in video sequences, but due to fast illumination change in the visual surveillance
system, many are not tolerant to dynamic background. In this paper, more focus is given for object detection using a pixel-
wise adaptive thresholding algorithm for an accurate detection of moving objects in video. The goal of this thresholding is to
classify pixels as either dark or light and this algorithm is more robust to illumination changes in the images. The proposed
algorithm is compared with existing object detection algorithms like object detection using simple background subtraction
using global thresholding(Otsu’s method) and with an self organizing approach to object detection using HSV color
representation model. The performance metrics used for comparing the results are Peak signal to noise ratio(PSNR), Root
mean square error(RMSE) and sensititivity and the results showed that proposed adaptive threshold based object detection
provides an accurate detection under varying illumination changes in the video sequence.