Motion correction method in dual energy radiography
A motion correction method includes two steps. The first step includes a global motion correction using the bilinear warping technique and a rough delineation of the lung fields. One of the native images (low energy image, high energy image) is deformed to match the other image. In a second step, local motion corrections are applied to the globally motion corrected image by computing a proximity value in small overlapping tiles. Only tiles with a sufficient high proximity value are taken into account. The maximum shift applied in this second step is limited to a few pixels to avoid strong deformations of the native images.
1. A method for correcting motion artifacts in a dual energy radiography image, the method comprising the steps of:
correcting a global position to globally correct positions of pixels in at least one of a high energy radiographic native image and a low energy radiographic native image of a same subject to obtain a globally corrected image; and
after the step of correcting the global position, correcting a local position to locally correct pixel positions in the globally corrected image; wherein
the step of correcting the global position includes:
determining, for pre-defined control points in one of the high energy radiographic native image and the low energy radiographic native image, a local proximity value for multiple translation offsets using local image patches centered at the pre-defined control points;
deriving a displacement vector for each of the pre-defined control points from the determined local proximity values;
deducing warping coefficients for a warping transformation from the displacement vectors; and
applying the warping transformation to one of the high energy radiographic native image and the low energy radiographic native image to obtain the globally corrected image; and
the step of correcting the local position of the globally corrected image includes:
defining overlapping tiles in the globally corrected image and a non-globally corrected high energy radiographic native image or a non-globally corrected low energy radiographic native image;
computing local displacement vectors for each of the overlapping tiles;
using the local displacement vectors to build a displacement map; and
applying the displacement map to the globally corrected image.
2. The method according to claim 1 , wherein the warping transformation is a bilinear warping transformation.
3. The method according to claim 1 , wherein the dual energy radiography image is a chest image and the control points are corner points of a quadrangle substantially delineating a lung field in the chest image.
4. The method according to claim 1 , further comprising defining and using additional control points to derive the displacement vector.
5. The method according to claim 1 , wherein the displacement vector is a translation offset for which the local proximity value obtained for the multiple translation offsets reaches a maximum and exceeds a predefined threshold.
6. The method according to claim 5 , wherein the local proximity value is a cross correlation coefficient of an image patch centered at a corner point in one of the high energy radiographic native image and the low energy radiographic native image and a template of the other of the high energy radiographic native image and the low energy radiographic native image.
7. The method according to claim 6 , wherein at least one of the high energy radiographic native image and the low energy radiographic native image is multiplicatively demodulated.
8. The method according to claim 5 , wherein the local proximity value is defined as a cross correlation coefficient of an image patch in the globally corrected image and a template of the non-globally corrected high energy radiographic native image or the non-globally corrected low energy radiographic native image.
9. The method according to claim 8 , wherein at least one of the globally corrected image and the non-globally corrected high energy radiographic native image or non-globally corrected low energy radiographic native image is multiplicatively demodulated.
10. The method according to claim 1 , wherein the local displacement vector for each pixel is computed by applying interpolation to values of the displacement map.
11. The method according to claim 1 , wherein the displacement map is applied to the globally corrected image using bilinear interpolation.
12. A non-transitory computer readable medium comprising computer executable program code adapted to carry out, when the computer executable program code is executed on a computer, the steps of:
correcting a global position to globally correct positions of pixels in at least one of a high energy radiographic native image and a low energy radiographic native image of a same subject to obtain a globally corrected image; and
after the step of correcting the global position, correcting a local position to locally correct pixel positions in the globally corrected image; wherein
the step of correcting the global position includes:
determining, for pre-defined control points in one of the high energy radiographic native image and the low energy radiographic native image, a local proximity value for multiple translation offsets using local image patches centered at the pre-defined control points;
deriving a displacement vector for each of the pre-defined control points from the determined local proximity values;
deducing warping coefficients for a warping transformation from the displacement vectors; and
applying the warping transformation to one of the high energy radiographic native image and the low energy radiographic native image to obtain the globally corrected image; and
the step of correcting the local position of the globally corrected image includes:
defining overlapping tiles in the globally corrected image and a non-globally corrected high energy radiographic native image or a non-globally corrected low energy radiographic native image;
computing local displacement vectors for each of the overlapping tiles;
using the local displacement vectors to build a displacement map; and
applying the displacement map to the globally corrected image.