Adaptive Lorentzian norm influence function for robust local based optical flow

Abstract

This paper introduces the using of adaptive Lorentzian norm influence function with the local based optical flow to achieve the robustness in noise interference on the image flow. In particularly, optical flow is the approach to approximate the motion flow from the image flow or video. When the image flow is contaminated or interfered by noise, it directly affects the performance of the optical flow to approximate the motion flow. We confirm better noise tolerance in local based optical flow when it is adapted to adaptive Lorentzian norm influence function in our experiment. We also prove the performance with several robust models when they are adapted for local based optical flow under the same environment. We simulate the additive white Gaussian (AWG) noise in several levels from low level of noise to high level of noise with several sets of image flow in the experiment where the peak signal to noise ratio (PSNR) is considered as the indication criteria.