Low Complexity Illumination-Invariant Motion Vector Detection Based on Logarithmic Edge Detection and Edge Difference

This paper describes a low complexity illumination-invariant motion vector detection algorithm based on logarithmic edge detection and edge difference without periodic threshold adjustment. A logarithmic edge detector is employed to achieve accurate object movement over a wide illumination range. The threshold for edge detection is determined by one-time background logarithmic gradient extraction. Finally, the difference of the logarithmic edge of 3 consecutive frames is employed in a frame difference-based motion template model to obtain the motion vector. Experimental results show that the proposed algorithm achieves a motion vector detection accuracy of ∼90% over an illumination level change of 80%.