Image.similarity#

Image.similarity(T: Any, metric: str = 'zncc') → Any[source]#

Locate template in image

Parameters:

  • T (ndarray(N,M)) – template image

  • metric (str) – similarity metric, one of: ‘ssd’, ‘zssd’, ‘ncc’, ‘zncc’ [default]

Raises:

  • ValueError – template T must have odd dimensions

  • ValueError – bad metric specified

Returns:

similarity image

Return type:

Image instance

Compute a similarity image where each output pixel is the similarity of the template T to the same-sized neighbourhood surrounding the corresponding input pixel in image.

Example:

>>> from machinevisiontoolbox import Image
>>> crowd = Image.Read("wheres-wally.png", mono=True, dtype="float")
>>> T = Image.Read("wally.png", mono=True, dtype="float")
>>> sim = crowd.similarity(T, "zncc")
>>> sim.disp(colormap="signed", colorbar=True);
<matplotlib.image.AxesImage object at 0x7f076d628e60>

Note

  • For NCC and ZNCC the maximum similarity value corresponds to the most likely template location. For SSD and ZSSD the minimum value corresponds to the most likely location.

  • Similarity is not computed for those pixels where the template crosses the image boundary, and these output pixels are set to NaN.

References:

Important

Uses OpenCV function cv2.matchTemplate which accepts multiple-channel, CV_8U or CV_32F images.

Seealso:

cv2.matchTemplate