System and method for counterfeit protection

• US 6,553,136 B1
• Filed: 10/28/1999
• Issued: 04/22/2003
• Est. Priority Date: 10/28/1999
• Status: Expired due to Fees

• Current Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Original Assignee: Hewlett-Packard Company (HP Inc.)
• Inventors: Shaked, Doron, Keshet, Renato
• Primary Examiner(s): Boudreau, Leo
• Assistant Examiner(s): Akhavannik, Hussein
• Application Number:
  US09/429,515
• Time in Patent Office: 1,272 Days
• Field of Search: 382/135
• US Class Current: 382/135
• CPC Class Codes: G06V 10/443 by matching or filtering
• CPC Class Codes: G06V 10/462 Salient features, e.g. scale invariant feature transforms [SIFT]
• CPC Class Codes: G06V 10/56 relating to colour

Abstract

A system and a method for preventing counterfeiting activities utilize an image detection scheme that uses only a few portions of restricted-document images, i.e., templates, to determine whether an input image is a duplicate of a restricted-document image. The image detection scheme can reliably operate at very low resolutions, which allows the use of small kernels for image correlation and reduces the number of pixels that have to be actually processed. The system is designed to be incorporated into a host device, such as a copy machine, a scanner, a personal computer or a printer, to detect a potential counterfeiting activity, so that an appropriate action can be taken to prevent and/or to report the detected activity. The image detection scheme allows the system to operate at an increased detection speed, while minimizing the memory requirement. During an off-line training procedure, a training module of the system derives a parameter file that contains detection descriptors for the restricted-document images, including threshold intervals that are defined using selected kernels and templates of the restricted-document images. The type of kernels and templates selected by the training module can vary, depending on the desired memory requirement and the desired on-line performance. During an on-line detection procedure, a detection module of the system projects portions of the input image and compares the resulting projection value with the threshold intervals to derive template match candidates. The relative coordinates of the template match candidates in the input image are then compared with relative coordinates of the templates in the restricted-document images to determine if the input image matches a particular restricted-document image.

Claims

1. A system for counterfeit protection comprising:

a host apparatus that processes an acquired input image;

training means for deriving parameters from a plurality of templates and kernels, said templates being selected portions of a restricted-document image, said kernels having a rotation-invariant convolution characteristic;

detecting means operatively coupled to said training means for determining whether said input image resembles said restricted-document image by utilizing said parameters, said detecting means being configured to compare relative positions of at least two template match candidates in said input image with relative positions of corresponding templates in said restricted-document image, each of said template match candidates being a portion of said input image that is preliminarily determined to correlate with one of said templates of said restricted-document image; and

signaling means operatively coupled to said detecting means for transmitting a signal when a positive determination is made by said detecting means that said input image closely resembles said restricted-document image.

2. The system of claim 1 wherein said comparing means is configured to compare relative coordinates of said template candidates in said input image with relative coordinates of said corresponding templates in said restricted-document image.

3. The system of claim 1 wherein said training means includes a processor for calculating said parameters from said plurality of said templates and kernels, said parameters being intervals of values that relate to an image characteristic of said templates.

4. The system of claim 1 wherein said training means includes a kernel selector and a template selector, said kernel and templates selectors being configured to select said kernels and said templates, respectively.

5. The system of claim 4 wherein said kernel selector of said training means is configured to select one of universal kernels and customized kernels, said universal kernels being independent of said templates, said customized kernels being dependent on said templates.

6. The system of claim 4 wherein said template selector of said training means is configured to select one of kernel-dependent templates and kernel-independent templates.

7. The system of claim 1 wherein said detecting means includes a linear match filter that operates to perform a convolution on said input image to produce projection values, said projection values being used to determine whether a particular portion of said input image is a template match candidate, said template match candidate being a portion of said input image that is preliminarily determined to correlate with one of said templates of said restricted-document image.

8. The system of claim 7 wherein said detecting means further includes a thresholder, said thresholder being configured to compare an image characteristic value of said particular portion of said input image with a corresponding parameter derived by said training means, said image characteristic value being related to said projection values produced by said linear match filter.

9. The system of claim 1 wherein said host apparatus being one of a copy machine, a scanner, a printer and a computer, said input image being a scanned image of a document generated by said host apparatus.

10. A method of detecting counterfeiting activities comprising steps of:

receiving an input image;

projecting portions of said input image with predefined kernels to produce projection values relating to at least one image characteristic, said predefined kernels having a rotation-invariant convolution property;

thresholding said projection values with predefined values that relate to said at least one image characteristic of a restricted-document image, so as to select template match candidates from said portions of said input image, said template match candidates being image portions of said input image that resemble at least one of templates of said restricted-document image;

comparing relative positions of said template match candidates in said input image with relative positions of corresponding templates in said restricted-document image to determine whether said input image is a duplicate of said restricted-document image; and

transmitting a signal when a determination is made that said input image is a duplicate of said restricted-document image.

11. The method of claim 10 further comprising steps of:

selecting said predefined kernels having said rotation-invariant convolution property;

selecting said templates extracted from said restricted-document image; and

computing said predefined values from said kernels and said templates.

12. The method of claim 11 wherein said step of selecting said predefined kernels includes a step of computing universal kernels, said universal kernels being independent of said templates.

13. The method of claim 11 wherein said step of selecting said templates includes a step of extracting said templates from said restricted-document image that are dependent on said predefined kernels.

14. The method of claim 11 wherein said step of computing said predefined values includes a step of defining an interval of values that relate to said at least one image characteristic of said templates of said restricted document.

15. The method of claim 10 further comprising a step of executing a preventive action in response to said signal in order to prevent counterfeiting of said restricted document.

16. A method of detecting counterfeiting activities comprising steps of:

deriving parameters from a plurality of templates and a plurality of kernels, said templates being selected portions of a restricted document that is to be protected from counterfeiting, said kernels having a rotation-invariant convolution characteristic; and

determining whether an input image closely resembles said restricted document by correlating portions of said input image with said templates of said restricted document using said kernels to derive projection values, including comparing said projection values with said parameters, said comparison of said projection values with said parameters being a factor for said determination of whether said input image closely resembles said restricted document, said determining including geometrically comparing relative positions of template match candidates in said input image with relative positions of said templates in said restricted document, each of said template match candidates being a particular portion of said input image that is preliminarily determined to correlate with one of said templates of said restricted document.

17. The method of claim 16 wherein said step of deriving said parameters includes a step of defining at least one interval of values that relates to an image characteristic of said templates of said restricted document.

18. The method of claim 17 wherein said step of defining said at least one interval is a step of defining said values so as to relate to color, intensity and angular orientation.

19. The method of claim 17 wherein said step of comparing said projection values includes a step of thresholding an image characteristic value with said interval, said image characteristic value being related to said projection values derived from said portions of said input image.

20. The method of claim 16 further comprising steps of:

selecting said kernels that are used in said step of deriving said parameters; and

selecting said templates that are used with said kernels in said step of deriving said parameters.

21. The method of claim 20 wherein said step of selecting said kernels includes a step of computing said kernels of a particular type selected from a group consisting of universal kernel and customized kernel, said universal kernels being independent of said templates, said customized kernels being dependent on said templates.

22. The method of claim 20 wherein said step of selecting said templates includes a step of extracting said templates from said restricted document, said templates being of a type selected from a group consisting of kernel-dependent templates and kernel-independent templates.

23. The method of claim 16 further comprising a step of transmitting a signal to a destination in response to a determination that said input image closely resembles said restricted document in order to prevent counterfeiting of said restricted document.

24. The method of claim 16 wherein said step of determining whether said input image closely resembles said restricted document is executed within an apparatus, said apparatus being one of a copy machine, a scanner, a printer and a computer.