Correcting registration errors in digital printing

1. A method for correcting an error in image printing, the method comprising:

receiving a reference digital image (RDI);

selecting, based on a predefined selection criterion, one or more regions in the RDI that are suitable for use as anchor features for sensing the error;

receiving a digital image (DI) acquired from a printed image of the RDI, and identifying the one or more regions in the DI;

estimating, based on the anchor features of the DI, the error in the printed image by training a neural network to detect the error in the one or more regions, and wherein training the neural network comprises producing multiple versions of the RDI having respective errors;

calculating a correction that, when applied to the DI, compensates for the estimated error;

correcting the estimated error in a subsequent digital image (SDI) to be printed; and

printing the SDI having the corrected error.

2. The method according to claim 1 , wherein the error comprises a registration error, and wherein training the neural network comprises comparing, in a selected region, between a predefined registration error produced in the selected region, and an output registration error estimated by the neural network, and applying, to the predefined and the output registration errors, a loss function, and comparing an output of the loss function to a threshold.

3. The method according to claim 1 , wherein the RDI comprises at least a first reference color image (RCI) having a first color, and a second RCI having a second color, and wherein producing the multiple versions comprises shifting, in the region, at least part of the first RCI relative to the second RCI.

4. The method according to claim 1 , wherein training the neural network comprises training a multi-layered convolutional neural network comprising at least five layers.

5. The method according to claim 1 , wherein the error comprises a registration error, and wherein estimating the registration error comprises estimating an image-to-substrate (ITS) registration error by (i) measuring, in the RDI, a distance between at least one of the anchor features and a given region at an edge of the printed image, and (ii) estimating, in the printed image of the DI, a variation of the distance.

6. A system for correcting an error in image printing, the system comprising:

a processor, which is configured to:

receive a reference digital image (RDI);

select, based on a predefined selection criterion, one or more regions in the RDI that are suitable for use as anchor features for sensing the error;

receive a digital image (DI) acquired from a printed image of the RDI, and identify the one or more regions in the DI;

estimate, based on the anchor features of the DI, the error in the printed image by training a neural network to detect the error in the one or more regions, and wherein the processor is configured to produce, for the training, multiple versions of the RDI having respective errors;

calculate a correction that, when applied to the DI, compensates for the estimated error; and

correct the estimated error in a subsequent digital image (SDI) to be printed; and

a printing subsystem configured to print the SDI having the corrected error.

7. The system according to claim 6 , wherein the error comprises a registration error, and wherein the processor is configured to: (i) compare, in a selected region, between a predefined registration error produced in the selected region, and an output registration error estimated by the neural network, (ii) apply, to the predefined registration error and to the output registration error, a loss function, and (iii) compare an output of the loss function to a threshold.

8. The system according to claim 6 , wherein the RDI comprises at least a first reference color image (RCI) having a first color, and a second RCI having a second color, and wherein the processor is configured to shift, in the region, at least part of the first RCI relative to the second RCI.

9. The system according to claim 6 , wherein the processor is configured to train a multi-layered convolutional neural network comprising at least five layers.

10. The system according to claim 6 , wherein the error comprises a registration error, and wherein the processor is configured to estimate an image-to-substrate (ITS) registration error, by (i) measuring, in the RDI, a distance between at least one of the anchor features and a given region at an edge of the printed image, and (ii) estimating, in the printed image of the DI, a variation of the distance.

11. A method for correcting a registration error in image printing, the method comprising:

receiving a reference digital image (RDI);

producing multiple versions of the RDI having respective registration errors, and selecting, based on a predefined selection criterion, one or more regions in the RDI that are suitable for use as anchor features for sensing the registration error, wherein producing the multiple versions comprises applying to at least part of at least one of the regions, one or more errors selected from a list consisting of: (a) flipping in one or more axes, (b) rotating in one or more axes, (c) image blurring, and (d) adding noise, and wherein selecting the one or more regions comprises training a neural network using the multiple versions;

receiving a digital image (DI) acquired from a printed image of the RDI, and identifying the one or more regions in the DI;

estimating, based on the anchor features of the DI, the registration error in the printed image;

calculating a correction that, when applied to the DI, compensates for the estimated registration error;

correcting the estimated registration error in a subsequent digital image (SDI) to be printed; and

printing the SDI having the corrected registration error.

12. A method for correcting a registration error in image printing, the method comprising:

receiving a reference digital image (RDI);

selecting, based on a predefined selection criterion, one or more regions in the RDI that are suitable for use as anchor features for sensing the registration error;

receiving a digital image (DI) acquired from a printed image of the RDI, and identifying the one or more regions in the DI;

estimating, based on the anchor features of the DI, the registration error in the printed image, by estimating, in at least one of the regions of the DI, a color-to-color (CTC) registration error between the first and second colors by (i) measuring, in at least one of the regions of the RDI image, a given distance between the first color and the second color, and (ii) estimating, in the printed image of the DI, a variation of the given distance;

calculating a correction that, when applied to the DI, compensates for the estimated registration error;

correcting the estimated registration error in a subsequent digital image (SDI) to be printed; and

printing the SDI having the corrected registration error.

13. A method for correcting an error in image printing, the method comprising:

receiving a reference digital image (RDI);

selecting, based on a predefined selection criterion, one or more regions in the RDI that are suitable for use as anchor features for sensing the error;

receiving a digital image (DI) acquired from a printed image of the RDI, and identifying the one or more regions in the DI;

estimating, based on the anchor features of the DI, the error in the printed image, wherein the error comprises at least one of: (a) a shift of a given pattern of the printed image relative to a position of the given pattern in the RDI, (b) a missing pattern, (c) a non-uniform thickness of a pattern in the printed image, (d) a non-uniform width of the pattern in the printed image, (e) a deviation in a profile of one or more colors of the printed image, and (f) a deviation in a linearity of one or more colors of the printed image;

calculating a correction that, when applied to the DI, compensates for the estimated error;

correcting the estimated error in a subsequent digital image (SDI) to be printed; and

printing the SDI having the corrected error.

14. A system for correcting a registration error in image printing, the system comprising:

a processor, which is configured to:

receive a reference digital image (RDI);

produce multiple versions of the RDI having respective registration errors, and select, based on a predefined selection criterion, one or more regions in the RDI that are suitable for use as anchor features for sensing the registration error, wherein the processor is configured to produce the multiple versions by applying to at least part of at least one of the regions, one or more errors selected from a list consisting of: (a) flipping in one or more axes, (b) rotating in one or more axes, (c) image blurring, and (d) adding noise, and wherein selecting the one or more regions comprises training a neural network using the multiple versions;

receive a digital image (DI) acquired from a printed image of the RDI, and identify the one or more regions in the DI;

estimate, based on the anchor features of the DI, the registration error in the printed image;

calculate a correction that, when applied to the DI, compensates for the estimated registration error; and

correct the estimated registration error in a subsequent digital image (SDI) to be printed; and

a printing subsystem, which is configured to print the SDI having the corrected registration error.

15. A system for correcting a registration error in image printing, the system comprising:

a processor, which is configured to:

receive a reference digital image (RDI);

select, based on a predefined selection criterion, one or more regions in the RDI that are suitable for use as anchor features for sensing the registration error;

receive a digital image (DI) acquired from a printed image of the RDI, and identify the one or more regions in the DI;

estimate, based on the anchor features of the DI, the registration error in the printed image, by estimating, in at least one of the regions of the DI, a color-to-color (CTC) registration error between the first and second colors by (i) measuring, in at least one of the regions of the RDI image, a given distance between the first color and the second color, and (ii) estimating, in the printed image of the DI, a variation of the given distance;

calculate a correction that, when applied to the DI, compensates for the estimated registration error; and

correct the estimated registration error in a subsequent digital image (SDI) to be printed; and

a printing subsystem, which is configured to print the SDI having the corrected registration error.

16. A system for correcting an error in image printing, the system comprising:

a processor, which is configured to:

receive a reference digital image (RDI);

select, based on a predefined selection criterion, one or more regions in the RDI that are suitable for use as anchor features for sensing the error;

receive a digital image (DI) acquired from a printed image of the RDI, and identify the one or more regions in the DI;

estimate, based on the anchor features of the DI, the error in the printed image, wherein the error comprises at least one of: (a) a shift of a given pattern of the printed image relative to a position of the given pattern in the RDI, (b) a missing pattern, (c) a non-uniform thickness of a pattern in the printed image, (d) a non-uniform width of the pattern in the printed image, (e) a deviation in a profile of one or more colors of the printed image, and (f) a deviation in a linearity of one or more colors of the printed image;

calculate a correction that, when applied to the DI, compensates for the estimated error; and

correct the estimated error in a subsequent digital image (SDI) to be printed; and

a printing subsystem, which is configured to print the SDI having the corrected error.