View-to-view image correction for object motion

Abstract

An imaging system of a type producing a plurality of X-Y matrices representing projection views of an object for later reconstruction using parts of a plurality of the views employs correlation techniques for determining the amount of motion along X and Y axes. The detected motion may be removed by suitably shifting later data to align it with earlier data, or vice versa. In one embodiment, all brightness values parallel to the X axes of two views are summed to develop two one-dimensional arrays containing information about the brightness patterns in the Y direction. One of the two one-dimensional arrays is shifted, and the two one-dimensional arrays multiplied, pixel-by-pixel, and the products are summed to provide a measure of similarity between the brightness patterns in the one-dimensional arrays. Repeated shifts produce a correlation function the shift position of whose maximum is a good approximation of the amount of motion occurring between the two views. The detected motion may be used to alert the operator that the data is faulty or to permit automatic alignment of all data from a scan. An identical process may be performed along the Y axis to detect and correct motion along the X axis, in certain restricted cases.