Wavelet Energy and the Usefulness of its Powers in Motion Detection

The potential for the usage of energy exponents in motion detection from video sequences is explored. The wavelet domain was chosen for the research due to the optimality of Hilbert's space for energy calculations and Parseval's equation for energy equivalence between domains. Five algorithms were considered: wavelet energy motion detection algorithm based on wavelet pairs and buffer, listed in the references, and four which are the contributions of this paper: modification by the application of different wavelet pairs, a modified algorithm without buffer, a modified algorithm without buffer and pairs, newly developed algorithms for energy exponents with and without buffer, but with wavelet pairs. The considered algorithms are background subtraction algorithms modified not to use pixels values, but rather energy/energy exponent backgrounds and the current situation models. These models are described by wavelet descriptors, the introduction of which is the contribution of this paper. They are compared by standard statistical criteria and execution time. The results suggest that an increase in the energy exponent decreases precision, recall and F-measure. However, the percentage of correct classifications remains almost constant. Higher exponentials reduce noise, but are more susceptible to shadows, the waving tree effect and similar abnormalities. Algorithms without buffers are less robust to illumination changes.

Abstract licence: CC BY-NC-ND

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.