Fig. 2.

Modeling spatially variant SR kernels. In this illustration, there are two regions (blue and white) having different LOR resolution in the radial direction. In each region, the ellipsoid SR kernel (red-dot boundary) shares the same location information (SR_loc) for the grid points within the kernel. For example, SR_loc will return (0, 1) for all points in the kernels (a, b, and c). This information can be used to fetch image data at those points by adding the origin of each kernel. However, each kernel can have a different SR_val according to its region. Assuming SR_loc is a following row-major order, for example, SR_val will return a′ and b′ for the points in a and b, respectively; but, for the point in c, SR_val will return c′ by fetching it in reverse order.