Moving Object Detection in Highly Corrupted Noise using Analysis of Variance

This paper implements three-way nested design to mark moving objects in a sequence of images. Algorithm performs object detection in the image motion analysis. The inter-frame changes (level-A) are marked as temporal contents, while the intra-frame variations identifies critical information. The spatial details are marked at two granular levels, comprising of level-B and level-C. The segmentation is performed using analysis of variance (ANOVA). This algorithm gives excellent results in situations where images are corrupted with heavy Gaussian noise ~N(0,100). The sample images are selected in four categories: ‘baseline’, ‘dynamic background’, ‘camera jitter’, and ‘shadows’. Results are compared with previously published results on four accounts: false positive rate (FPR), false negative rate (FNR), percentage of wrong classification (PWC), and an F-measure. The qualitative and quantitative results prove that the technique out performs the previously reported results by a significant margin.