Data processing apparatus and method

• US 20060126892A1
• Filed: 12/07/2005
• Published: 06/15/2006
• Est. Priority Date: 12/09/2004
• Status: Active Grant

• Current Assignee: Sony United Kingdom Limited (Sony Group Corp.)
• Original Assignee: Sony United Kingdom Limited (Sony Group Corp.)
• Inventors: Tapson, Daniel Warren
• Granted Patent: US 7,609,850 B2
• Time in Patent Office: Days
• US Class Current: 382/100
• CPC Class Codes: G06T 1/005 Robust watermarking, e.g. average attack or collusion attack resistant
• CPC Class Codes: G06T 2201/0051 Embedding of the watermark in the spatial domain
• CPC Class Codes: G06T 2201/0061 Embedding of the watermark in each block of the image, e.g. segmented watermarking
• CPC Class Codes: G06T 2201/0065 Extraction of an embedded watermark; Reliable detection
• CPC Class Codes: G06T 2201/0081 whereby both original and watermarked images are required at decoder, e.g. destination-based, non-blind, non-oblivious
• CPC Class Codes: G06T 7/0002 Inspection of images, e.g. flaw detection
• CPC Class Codes: H04N 1/32154 Transform domain methods H04N1/32309 takes precedence
• CPC Class Codes: H04N 2201/3233 of authentication information, e.g. digital signature, watermark
• CPC Class Codes: H04N 2201/327 which are undetectable to the naked eye, e.g. embedded codes

Abstract

A data processing apparatus detects a code word, which has been embedded in a water marked image. The apparatus recovers a code word estimate from the water marked image and correlates the estimated code word with each code word from a set of possible code words. The apparatus detects whether a code word is present from a result of the correlation. In order to improve a likelihood of correctly detecting a code word, which may be present in a water marked image, the data processing apparatus includes a registration processor. The registration processor includes an image analyser operable to divide the water marked image into a plurality of image blocks, a vector estimator, a forward probability estimator and a backward probability estimator. The vector estimator generates for each of the image blocks a probability distribution of distortion vectors representing an observed shift of the image block with respect to a position of the block in the original version of the image. The forward probability estimator generates for each of the image blocks a forward probability distribution of possible distortion vectors within the image block from the observed probability distribution of distortion vectors and the distortion probabilities from previous image blocks following a predetermined path through the blocks. The backward probability estimator generates for each of the image blocks a backward probability distribution of possible distortion vectors within the image block from the observed distortion probabilities for the image block and from subsequent image blocks following the predetermined pattern. A combining engine determines a most likely distortion vector for each image block from a combination of the estimated distortion vector, the forward probability estimate and the backward probability estimate for each image block. An inverse distortion engine applies a reverse of the most likely distortion vector for each image block to register the water marked image with the original image to improve a likelihood of correctly detecting the water mark code word.

Claims

1. A data processing apparatus for detecting at least one of a set of code words which has been embedded in a water marked image, the apparatus comprising

a registration processor operable to register the water marked image and an original version of the image with respect to one another, and

a detection processor operable

to recover an estimate of the code word from the water marked image by comparing the registered water marked image with the original version of the image, and

to detect the code word by correlating each of the code words of the set with the estimated code word, wherein the registration processor includes

an image analyser operable to divide a frame of the water marked image into a plurality of image blocks,

a distortion vector probability estimator operable to estimate for each of the image blocks of the water marked image an observed probability of possible distortion vectors representing shifts of the image block with respect to positions of the block in the original version of the image,

a forward probability estimator operable for each of the image blocks to generate an estimate of a forward probability distribution of possible distortion vectors within the image block recursively from the observed probability of distortion vectors for the image block and forward probability distribution estimates determined from previous image blocks for a predetermined path through the blocks,

a backward probability estimator operable for each of the image blocks to generate an estimate of a backward probability distribution of possible distortion vectors within the image block recursively from the observed probability of distortion vectors for the image block and backward probability distribution estimates determined from subsequent image blocks for the predetermined pattern, and

a combining engine operable to combine for each image block the forward probability distribution estimate, the backward probability distribution estimate and the observed probability of distortion vector to form a most likely distortion vector estimate, and

to form a map of distortion vectors from the most likely distortion vector estimate for each image block, and

an inverse distortion engine operable to apply a reverse of the map of most likely distortion vectors to register the water marked image with the copy of the original image to improve a likelihood of correctly detecting the water mark code word.

2. A data processing apparatus as claimed in claim 1, wherein the inverse distortion engine comprises

a modelling processor operable to generate a modelling function which best fits the map of the most likely distortion vector probabilities, the modelling function representing distortion present in the water marked image, and

a re-sampler operable to re-sample the water marked image to an effect of reducing distortion in the water marked image represented by the modelling function to register the water marked image with the original version of the image.

3. A data processing apparatus as claimed in claim 2, wherein the re-sampler is operable to re-sample the pixels of the image to reverse the distortion according to the modelling function.

4. A data processing apparatus as claimed in claim 1, wherein the predetermined pattern comprises first and second predetermined patterns and the forward and backward probability estimators are first forward and backward probability estimators, the apparatus comprising

a second forward probability estimator operable for each of the image blocks to generate a second estimate of the forward probability distribution of possible distortion vectors within the image block recursively from the most likely distortion vectors for the image block and the forward probability distribution estimates determined from previous image blocks for the second predetermined path through the blocks,

a second backward probability estimator operable for each of the image blocks to generate a second estimate of a backward probability distribution of possible distortion vectors within the image block recursively from the most likely distortion vectors for the image block and the backward probability distribution estimates determined from subsequent image blocks for the second predetermined pattern, wherein the combining engine is operable to combine the second forward probability distribution estimates, with the second backward probability distribution estimates and the most likely distortion vectors to form a refined distortion probability map providing a refinement of the most likely distortion vectors, the inverse distortion engine being operable to apply a reverse of the refined distortion probability map to reduce distortion in the water marked image to register the water marked image with the copy of the original image.

5. A data processing apparatus as claimed in claim 1, wherein the first predetermined pattern identifies the image blocks for processing row-by-row, and the second predetermined pattern identifies the image blocks for processing column-by-column.

6. A data processing apparatus as claimed in claim 1, wherein at least one of the first and second forward probability estimators includes

a filter for adjusting the forward probability distribution estimates of distortion vectors in accordance with a predetermined probability distribution effecting a likelihood of the distortion vector for the forward direction occurring, and at least one of the first and second backward probability estimators includes

a filter for adjusting the backward probability distribution of distortion vectors in accordance with the predetermined probability distribution effecting a likelihood of the distortion vector for the backward direction occurring.

7. A data processing apparatus as claimed in claim 6, wherein the predetermined probability distribution is a Gaussian distribution.

11. A method of detecting as claimed in claim 7, wherein the predetermined pattern comprises first and second predetermined patterns, the method comprising

generating for each of the image blocks a second estimate of the forward probability distribution of possible distortion vectors within the image block from the most likely distortion vector for the image block and the forward probability distribution estimates determined from previous image blocks for the second predetermined path through the image blocks,

generating for each of the image blocks a second estimate of a backward probability distribution of possible distortion vectors within the image block from the most likely distortion vector for the image block and the backward probability distribution estimates determined from subsequent image blocks for the second predetermined pattern, and

combining the second forward probability distribution of distortion vectors, with the second backward probability distribution of distortion vectors and the most likely distortion vectors to form a refined distortion probability map providing a refinement of the most likely distortion vectors, and

applying a reverse of the refined distortion probability map to reduce distortion in the water marked image so as to register the water marked image with the original version of the image.

12. A method of detecting as claimed in claim 7, wherein the first predetermined pattern identifies the image blocks for processing row-by-row, and the second predetermined pattern identifies the image blocks for processing column-by-column.

13. A method as claimed in claim 7, comprising

adjusting the forward probability distribution estimates of distortion vectors in accordance with a predetermined probability distribution effecting a likelihood of the distortion vector for the forward direction occurring,

adjusting the backward probability distribution estimates of distortion vectors in accordance with the predetermined probability distribution effecting a likelihood of the distortion vector for the backward direction occurring.

14. A method as claimed in claim 13, wherein the predetermined probability distribution is a Gaussian distribution.

8. A method of detecting at least one of a set of code words which has been embedded in a water marked image, the method comprising

registering the water marked image and an original version of the image with respect to one another,

recovering an estimate of the code word from the water marked image by comparing the registered water marked image with the original version of the image, and

detecting the code word by correlating each of the plurality of code words of the set with the estimated code word, wherein the registering the water marked image and the original version of the image, comprises

dividing a frame of the water marked image into a plurality of image blocks,

estimating for each of the image blocks of the water marked image an observed probability of possible distortion vectors representing shifts of the image block with respect to positions of the block in the original version of the image,

generating for each of the image blocks an estimate of a forward probability distribution of possible distortion vectors within the image block recursively from the observed probability of distortion vectors for the image block and forward probability distribution estimates determined from previous image blocks for a predetermined path through the blocks,

generating for each of the image blocks an estimate of a backward probability distribution of possible distortion vectors within the image block recursively from the observed probability of distortion vectors for the image block and backward probability distribution estimates determined from subsequent image blocks for the predetermined pattern, and

combining for each image block the forward probability distribution estimate, the backward probability distribution estimate and the observed probability of distortion vector to form a most likely distortion vector estimate, and

forming a map of distortion vectors from the most likely distortion vector estimate for each image block, and

applying a reverse of the most likely distortion vector for each image block by re-sampling the image to register the water marked image with the original version of the image to improve a likelihood of correctly detecting the water mark code word.

9. A method of detecting as claimed in claim 8, wherein the applying a reverse of the most likely distortion vector comprises

generating a modelling function which best fits the map of the most likely distortion vector probabilities, the modelling function representing distortion present in the water marked image, and

re-sampling the water marked image to an effect of reducing distortion in the water marked image represented by the modelling function to register the water marked image with the original version of the original image.

10. A method of detecting as claimed in claim 9, wherein the re-sampling comprises

re-sampling the pixels of the image to reverse the distortion according to the modelling function.

15. A computer program providing computer executable instructions, which when loaded on to a data processor causes the data processor to perform the method according to claim 8.

16. A computer program product having a computer readable medium having recorded thereon information signals representative of the computer program claimed in claim 15.