Registration of histopathology images

1. An imaging processing component of an image processing system, the imaging processing component comprising:

an input interface to receive at least two sets of image objects, wherein at least one set includes a single image object or a plurality of image objects, said at least two sets of image objects derived from imagery of biological material;

an image feature deriver configured to derive image features of or in relation to the at least two sets of image objects;

a global transformation builder configured to compute, based on said image features, an initial global registration transformation for the at least two sets of image objects;

a cross-linker configured to establish, based on said initial global registration transformation and its inverse, at least one cross-link between image objects from the at least two sets of image objects, wherein the at least one cross-link is based on mapping action points in one of the at least two sets of image objects onto another of the at least two sets of image objects using the initial global registration transformation and then mapping action points in the another of the at least two sets of image objects onto the one of the at least two sets of image objects using the inverse of the initial global registration transformation;

wherein each of the image objects is a region or set of image points of an image, and

wherein a cross-link lumps together all those image objects across the at least two sets of image objects that can be reached by a mapping action of the initial global registration transformation or its inverse.

2. The image processing component of claim 1 , further comprising:

a local transformation builder, configured to compute for said cross-link, at least one locally adapted transformation that is restricted to those image objects that are cross-linked in said cross-link;

a combiner configured to perform a combination operation to combine i) at least said locally adapted transformation and at least a part of said initial global registration transformation or ii) a plurality of locally adapted transformations into a combined registration transformation for the at least two sets of image objects; and

an output interface to output said new global transformation.

3. The image processing component of claim 2 , wherein operation of the local transformation builder for computing the at least one locally adapted transformation is based on one or more of the features previously derived by the image feature deriver.

4. The image processing component of claim 1 , comprising a refiner configured to:

sub-divide respective image objects from said at least two sets of image objects, into sub-objects to obtain sets of sub-objects;

compute, for said sub-objects at least one sub-locally adapted transformation; wherein the combiner is operative to include into said combination operation, said at least one sub-locally adapted transformation.

5. The image processing component of claim 1 , comprising a graphics display generator operative to one or a combination of effect displaying a registration result as per said initial global transformation or to update the displaying for a new registration result as per an intermediate registration or as per the combined registration transformation.

6. The image processing component of claim 5 , further including a sampler configured to up- or down-sample at least one of the at least two sets of image objects in case no initial global transformation can be computed by the global transformation builder.

7. The image processing component of claim 2 , wherein operation of the combiner is based on B-splining.

8. The image processing component of claim 1 , wherein at least one of said at least two sets of image objects has been derived by an image segmentation operation applied to said imagery.

9. An image processing system comprising:

an image processing component, comprising:

an input interface to receive at least two sets of image objects, wherein at least one set includes a single image object or a plurality of image objects, said at least two sets of image objects derived from imagery of biological material;

an image feature deriver configured to derive image features of or in relation to the at least two sets of image objects;

a global transformation builder configured to compute, based on said image features, an initial global registration transformation for the at least two sets of image objects;

a cross-linker configured to establish, based on said initial global registration transformation and its inverse, at least one cross-link between image objects from the at least two sets of image objects, wherein the at least one cross-link is based on mapping action points in one of the at least two sets of image objects onto another of the at least two sets of image objects using the initial global registration transformation and then mapping action points in the another of the at least two sets of image objects onto the one of the at least two sets of image objects using the inverse of the initial global registration transformation;

wherein each of the image objects is a region or set of image points of an image, and

wherein a cross-link lumps together all those image objects across the at least two sets of image objects that can be reached by a mapping action of the initial global registration transformation or its inverse; and

an image capturing device supplying the imagery from which the at least two sets of image objects have been derived.

10. An image processing method, comprising:

receiving at least two sets of image objects, wherein at least one set includes a single image object or a plurality of image objects, said at least two sets of image objects derived from imagery of biological material;

deriving image features of or in relation to the image object in the at least two sets of image objects;

based on said image features, computing an initial global registration transformation for the at least two sets of image objects;

establishing, based on said initial global registration transformation and its inverse, at least one cross-link between image objects from the at least two sets of image objects, wherein the at least one cross-link is based on mapping action points in one of the at least two sets of image objects onto another of the at least two sets of image objects using the initial global registration transformation and then mapping action points in the another of the at least two sets of image objects onto the one of the at least two sets of image objects using the inverse of the initial global registration transformation.

11. The image processing method of claim 10 , comprising:

for said at least one cross-link, computing at least one locally adapted transformation that is restricted to those image objects that are cross-linked in said cross-link;

combining i) at least said locally adapted transformation and at least a part of said initial global registration transformation or ii) a plurality of locally adapted transformations into a combined registration transformation for the at least two sets of image objects; and

outputting said combined transformation.

12. The image processing method of claim 11 , wherein the step of computing the at least one locally adapted transformation is based on one or more of the features previously derived at the deriving step.

13. The image processing method of claim 10 , wherein said imagery of biological material includes at least one digital pathology image.

14. The image processing method of claim 10 , wherein the computing of the at least one locally adapted function is performed at a same resolution level than the computing of the initial global registration transformation.

15. A computer program element for controlling a component, which, when being executed by a processing unit is adapted to perform the method steps of claim 10 .

16. A non-transitory computer readable medium having stored thereon the computer program element of claim 15 .