Abstract

Methods: Conventional edge detection algorithms (eg, Canny [2]) use magnitude images as the inputs and convolve them with gradient masks to determine edge strength and direction. In this work, we perform the gradient operation in Fourier domain, and use that as a guide for appropriate sub4sampling of k4space. The acquired k4space data is directly multiplied by the Fourier transform (FT) of x and y gradient masks. Fig. 1 contains 1D magnitude plots of the data, filters, and product of the two. Note that most of the energy in k4space is concentrated around the k4space origin. After multiplication with gradient filters in k4space, the distribution describes the energy distribution of the edge information. We define sub4sampling schemes that only consider k4space samples where the data is expected to be above a pre4defined threshold, ie, preserving the k4 space region where the most of the edge energy is …

Date

2007

Authors

Y Kim, J Nielsen, S Narayanan, K Nayak

Journal

Proc. 15th Annu. Meeting Int. Society Magnetic Resononace in Medicine

Pages

3458