Contour tracking on sequences of ventriculographic images: a comparison between gradient of Gaussian and first order absolute moment

In this paper the properties of both the GoG and FOAM operators, as well as the results obtained with their respective localization procedures, are compared. FOAM (first order absolute moment) and GoG (gradient of Gaussian) filters are used to enhance the discontinuities between the gray levels of the different structures of the images. The contour we are tracking is then localized through the filtered images by starting from an approximate contour. Results of the two automatic contour tracking procedures are presented on sequences of ventriculographic and synthetic images.