Finds the geometric transform (warp) between two images in terms of the ECC criterion

warpMatrix = cv.findTransformECC(templateImage, inputImage)
[warpMatrix,rho] = cv.findTransformECC(templateImage, inputImage)
[...] = cv.findTransformECC(..., 'OptionName',optionValue, ...)

Input

• templateImage single-channel template image; uint8 or single array.
• inputImage single-channel input image which should be warped with the final warpMatrix in order to provide an image similar to templateImage, same type as templateImage.

Output

• warpMatrix floating-point 2x3 or 3x3 mapping matrix (warp).
• rho correlation coefficient.

Options

• MotionType parameter, specifying the type of motion:
  • Translation sets a translational motion model; warpMatrix is 2x3 with the first 2x2 part being the unity matrix and the rest two parameters being estimated.
  • Euclidean sets a Euclidean (rigid) transformation as motion model; three parameters are estimated; warpMatrix is 2x3.
  • Affine sets an affine motion model (DEFAULT); six parameters are estimated; warpMatrix is 2x3.
  • Homography sets a homography as a motion model; eight parameters are estimated; warpMatrix is 3x3.
• Criteria parameter, specifying the termination criteria of the ECC algorithm; Criteria.epsilon defines the threshold of the increment in the correlation coefficient between two iterations (a negative Criteria.epsilon makes Criteria.maxcount the only termination criterion). Default values are: struct()
• Mask An optional mask to indicate valid values of inputImage. Not set by default.
• InputWarpMatrix Initial estimate for warpMatrix. Not set by default

The function estimates the optimum transformation (warpMatrix) with respect to ECC criterion ([EP08]), that is:

warpMatrix = argmax_{W} ECC(templateImage(x,y), inputImage(x',y'))

where:

[x';y'] = W * [x;y;1]

(the equation holds with homogeneous coordinates for homography). It returns the final enhanced correlation coefficient, that is the correlation coefficient between the template image and the final warped input image. When a 3x3 matrix is given with MotionType being one of 'Translation', 'Euclidean', or 'Affine', the third row is ignored.

Unlike cv.findHomography and cv.estimateRigidTransform, the function cv.findTransformECC implements an area-based alignment that builds on intensity similarities. In essence, the function updates the initial transformation that roughly aligns the images. If this information is missing, the identity warp (unity matrix) should be given as input. Note that if images undergo strong displacements/rotations, an initial transformation that roughly aligns the images is necessary (e.g., a simple euclidean/similarity transform that allows for the images showing the same image content approximately). Use inverse warping in the second image to take an image close to the first one, i.e. use the flag WarpInverse with cv.warpAffine or cv.warpPerspective. Note that the function throws an exception if algorithm does not converges.

References

[EP08]:

Georgios D Evangelidis and Emmanouil Z Psarakis. "Parametric image alignment using enhanced correlation coefficient maximization". Pattern Analysis and Machine Intelligence, IEEE Transactions on, 30(10):1858-1865, 2008.