An error analysis of camera translation direction estimation from optical flow using linear.pdfVIP

AN ERROR ANALYSIS OF CAMERA TRANSLATION DIRECTION ESTIMATION FROM OPTICAL FLOW USING LINEAR CONSTRAINTS A Mark Earnshaw and Steven D Blostein Department of Electrical and Computer Engineering Queens University Kingston Ontario KL N Canada ABSTRACT for a noisy constraint represented a attened ellipsoi d i A linear metho d using optical ow to recover the translation rather than a sphere They then prop osed the following direction of a moving camera has b een previously presented antibiasing technique This algorithm pro duced a biased estimate and various com p ensation techniques have b een prop osed The results of an Dithering Metho d error analysis of the dierent approaches containing expres The attened ellipsoi d nature of can b e observed by i sions for the exp ected mean vectors and covariance matrices considering its eigenvalues and eigenvectors The minimum of the translation estimates are presented here This anal eigenvalue eigenvector represents the mean sampling direc ysis includes a necessary mo dication to the CramerRao tion for the convolution patch which pro duced i Thus by lower b ound which sp ecies the minimum p ossible covari adding noise dithering in this direction the resulting co ance matrix for the dierent estimators The analytical variance matrix will approximate a sphere and the bias will results were veried exp erimentall y using simulations and b e eliminated The exp ected value of the dithered least real images squares matrix can th