Signal processors and methods for estimating transformations between signals with phase deviation
A phase deviation method determines an offset between a reference and suspect signal by analyzing a phase deviation surface created by computing a deviation metric for phase shift and then analyzing a surface formed from the deviation metrics for an array of offsets. The phase deviation method analyzes the deviation surface to determine an offset that minimizes phase deviation. This method is applied at increasing levels of detail to refine the determination of the offset.
1. A method of determining a geometric transformation of a reference signal in an image, the method comprising:
providing a linear transform candidate;
transforming a reference signal with the linear transform candidate to determine locations within the image;
measuring phases of the image at the locations to provide measured phase differences relative to known phases of the reference signal;
for each of plural candidate offsets, obtaining expected phase differences of the reference signal at the candidate offset and computing phase deviations for the expected and corresponding measured phase differences, each of the phase deviations comprising a deviation between an expected and corresponding measured phase difference;
summing the phase deviations for each candidate offset to provide an array of sums corresponding to the candidate offsets; and
searching the array of sums to find one or more offset candidates of the array that best approximate the geometric transformation of the reference signal in the image.
2. The method of claim 1 wherein the reference signal comprises a set of signal components, and the expected phase differences correspond to phases of the signal components for a candidate offset.
3. The method of claim 2 wherein the signal components are frequency components and the expected phase differences correspond to phase values of the frequency components after a geometric transformation and at a candidate offset.
4. The method of claim 1 wherein the image includes a version of the reference signal, and the sums are searched to determine the geometric transformation of the version of the reference signal in the image.
5. The method of claim 4 wherein the candidate offset together with rotation, scale, differential scale and shear parameters of the linear transform are used to determine the geometric transformation of the version of the reference signal in the image.
6. The method of claim 4 wherein the reference signal corresponds to a digital watermark signal embedded in the image, and the geometric transformation is used to detect the digital watermark signal in the image.
7. The method of claim 1 wherein the sums are searched to find a peak, which is then evaluated relative to a threshold to classify the image as including the reference signal.
8. The method of claim 1 wherein the sums are searched to detect presence of a watermark signal.
9. The method of claim 1 wherein the reference signal is constructed to have symmetry about an axis, and the reference signal has components that cancel each other when combined in one dimension to create a first 1D signal in a orthogonal dimension; and
wherein the array comprises plural 1 dimensional arrays of sums for candidate offsets for the first 1D signal, and the searching comprises searching the 1 dimensional arrays for peaks to get offset candidates for the first 1D signal.
10. A non-transitory computer readable medium, on which is stored instructions, which, when executed by one or more processors, perform a method of determining a geometric transformation of a reference signal in an image, the method comprising:
providing a linear transform candidate;
transforming a reference signal with the linear transform candidate to determine locations within the image;
measuring phases of the image at the locations to provide measured phase differences relative to known phases of the reference signal;
for each of plural candidate offsets, obtaining expected phase differences of the reference signal at the candidate offset and computing phase deviations for the expected and corresponding measured phase differences, each of the phase deviations comprising a deviation between an expected and corresponding measured phase difference;
summing the phase deviations for each candidate offset to provide an array of sums corresponding to the candidate offsets; and
searching the array of sums to find one or more offset candidates of the array that best approximate the geometric transformation of the reference signal in the image.
11. The non-transitory computer readable medium of claim 10 wherein the reference signal comprises a set of signal components, and the expected phase differences correspond to phases of the signal components for a candidate offset.
12. The non-transitory computer readable medium of claim 11 wherein the signal components are frequency components and the expected phase differences correspond to phase values of the frequency components after a geometric transformation and at a candidate offset.
13. The non-transitory computer readable medium of claim 10 wherein the image includes a version of the reference signal, and the sums are searched to determine the geometric transformation of the version of the reference signal in the image.
14. The non-transitory computer readable medium of claim 13 wherein the candidate offset together with rotation and scale parameters of the linear transform are used to determine the geometric transformation of the version of the reference signal in the image.
15. The non-transitory computer readable medium of claim 13 wherein the reference signal corresponds to a digital watermark signal embedded in the image, and the geometric transformation is used to detect the digital watermark signal in the image.
16. The non-transitory computer readable medium of claim 10 wherein the sums are searched to find a peak, which is then evaluated relative to a threshold to classify the image.
17. The non-transitory computer readable medium of claim 10 wherein the sums are searched to detect presence of a watermark signal.
18. The non-transitory computer readable medium of claim 10 wherein the reference signal is constructed to have symmetry about an axis, and the reference signal has components that cancel each other when combined in one dimension to create a first 1D signal in a orthogonal dimension; and
wherein the array comprises plural 1 dimensional arrays of sums for candidate offsets for the first 1D signal, and the searching comprises searching the 1 dimensional arrays for peaks to get offset candidates for the first 1D signal.
19. A circuit comprising:
a linear transform estimation module for estimating a linear transform of a reference signal in an image;
a transform coordinate module for transforming the reference signal with the linear transform candidate to determine locations within the image;
a phase estimation module for measuring phases of the image at the locations to provide measured phases;
a phase deviation module for generating an array of phase deviation sums corresponding to the measured phases for each of a set of candidate offsets, the phase deviation sums each comprising a sum of deviations between measured phase differences and expected phase differences at a candidate offset; and
a peak determination module for searching the array of sums to find one or more top offset candidates of the array that approximate the geometric transformation of the reference signal in the image.
20. The circuit of claim 19 wherein the linear transform estimation module is configured to determine the linear transform by finding a least squares fit between components of the reference signal and image features detected in the image.