Image registration

1. An image registration method for registering a target image with a reference image, comprising:

determining a target set of features present in the target image;

determining a reference set of features present in the reference image;

generating respective signatures for each feature in the target and reference sets;

matching features from the target set to corresponding features in the reference set using the signatures;

generating respective transformations to map pairs of features in the target set to corresponding pairs of features in the reference set; and

determining a global transformation for the target image using a measure of the accuracy of the transformations,

wherein determining target and reference sets of features includes determining regions with similar colors in the target and reference images and clustering such regions, each cluster representing a feature centered on the cluster, wherein the center is described by both its color and position.

2. An image registration method as claimed in claim 1 , wherein a signature for a feature is a geometric signature including a histogram of normalized distances between the feature and all other features in the set.

3. An image registration method as claimed in claim 1 , wherein transformations are geometric transformations generated over all pairs of determined image features.

4. An image registration method as claimed in claim 1 , wherein generating transformations to map pairs of features further comprises:

constructing a line segment between a pair of features in the target set;

constructing a line segment between a corresponding pair of features in the reference set, and determining a transformation to transform the line segment from the pair of features in the target set to the line segment from the pair of features in reference set, the transformation including at least one measure for an angle of rotation, translation, or scale factor; and

repeating over all pairs of determined image features.

5. An image registration method as claimed in claim 1 , wherein the global transformation for the target image is the transformation which occurs most often.

6. An image registration method as claimed in claim 5 , further comprising:

determining the number of occurrences of a transformation, wherein each occurrence increases the rank of a transformation over all transformations.

7. An image registration method as claimed in claim 6 , wherein determining the global transformation includes selecting the transformation with the highest rank.

8. A print system, comprising:

a print assembly;

an image processor to convert an input image to a raster image for the print assembly and to provide a reference image from the input image or raster image;

an image capture assembly to capture an image of at least a portion of a printed medium in order to provide a target image; and

a registration module to

generate a transformation for transforming the target image for registration with the reference image by matching features in the target image with equivalent features in the reference image using signatures for respective ones of the features,

determine a global transformation for the target image based on the transformation of pairs of features in the target image to equivalent pairs of features in the reference image, wherein to determine equivalent features the registration module is to

determine features in the target and reference images according to a measure of clusters of colors in the images, and

generate signatures for respective ones of the determined features, wherein a signature for a feature is a geometric signature including a histogram of normalized distances from the feature to all other determined features for the image.

9. A print system as claimed in claim 8 , wherein the registration module is further to register the target image with the reference image using the global transformation and to detect defects in the target image by comparing the registered images.

10. A print system as claimed in claim 8 , wherein to determine a global transformation for the target image the registration module is to:

construct a line segment between a pair of features in the target image;

construct a line segment between an equivalent pair of features in the reference image and determine a geometric transformation to transform the line segment of the target image to the line segment of the reference image, the transformation including at least one measure for an angle of rotation, translation, and scale factor.

11. A print system as claimed in claim 8 , wherein the registration module is further to arrange signatures for the target and reference images into respective sets, and to form a maximum weighted bipartite matching between signatures in the sets.

12. A non-transitory computer readable medium including machine readable instructions for instructing a print system to perform an automated method including:

to provide a reference image, at least a portion of which for forming the basis of a printed image;

to process a raster image;

to send the raster image to a print process;

to print a printed image corresponding to the raster image onto a medium;

to capture a target image from at least a part of the printed image, and

to determine a global feature-based transform for the target image based on feature matching between the target and reference images, and

to register the reference image and the target image using the determined transform,

wherein features are determined according to a measure of clusters of colors in the target and reference images, each cluster representing a feature centered on the cluster, wherein the center is described by both its color and position.

13. A computer readable medium as claimed in claim 12 including machine readable instructions for instructing a print system to perform an automated method further including:

to determine respective transforms for transforming line segments disposed between pairs of features in the target image to line segments disposed between corresponding pairs of features in the reference image.