Relative Velocity Estimation from a single Uniform Linear motion Blurred Image using The Discrete Cosine TransformCP5

Motion blur usually happens to image recording systems. Nowadays, this rich source of data is mostly dismissed in favor of image restoration. We introduce a novel approach for estimating kinematic quantities such as direction and speed of motion using the Discrete Cosine Transform (DCT). The mean relative error of our DCT PSeudo Cepstrum for speed estimation was 5.15%. The DCT frequency analysis is more accurate than all competitors evaluated.

This presentation is part of Contributed Presentation “CP5 - Contributed session 5

Jimy Alexander Cortes (Universidad Tecnologica de Pereira)
Juan Bernardo Gómez (Universidad Nacional de Colombia)
Juan Carlos Riaño (Universidad Nacional de Colombia)
cepstrum, computer vision, direction, image deblurring, image reconstruction, kinematic quantities, linear motion blur, speed