dsift¶
-
menpo.feature.
dsift
(image, *args, **kwargs)[source]¶ Computes a 2-dimensional dense SIFT features image with
C
number of channels, whereC = num_bins_horizontal * num_bins_vertical * num_or_bins
. The dense SIFT [2] implementation is taken from Vlfeat [1].Parameters: - pixels (
Image
or subclass or(C, Y, X)
ndarray) – Either the image object itself or an array with the pixels. The first dimension is interpreted as channels. - window_step_horizontal (int, optional) – Defines the horizontal step by which the window is moved, thus it controls the features density. The metric unit is pixels.
- window_step_vertical (int, optional) – Defines the vertical step by which the window is moved, thus it controls the features density. The metric unit is pixels.
- num_bins_horizontal (int, optional) – Defines the number of histogram bins in the X direction.
- num_bins_vertical (int, optional) – Defines the number of histogram bins in the Y direction.
- num_or_bins (int, optional) – Defines the number of orientation histogram bins.
- cell_size_horizontal (int, optional) – Defines cell width in pixels. The cell is the region that is covered by a spatial bin.
- cell_size_vertical (int, optional) – Defines cell height in pixels. The cell is the region that is covered by a spatial bin.
- fast (bool, optional) – If
True
, then the windowing function is a piecewise-flat, rather than Gaussian. While this breaks exact SIFT equivalence, in practice it is much faster to compute. - verbose (bool, optional) – Flag to print SIFT related information.
Raises: ValueError
– Only 2D arrays are supportedValueError
– Size must only contain positive integers.ValueError
– Step must only contain positive integers.ValueError
– Window size must be a positive integer.ValueError
– Geometry must only contain positive integers.
References
[1] Vedaldi, Andrea, and Brian Fulkerson. “VLFeat: An open and portable library of computer vision algorithms.” Proceedings of the international conference on Multimedia. ACM, 2010. [2] Lowe, David G. “Distinctive image features from scale-invariant keypoints.” International journal of computer vision 60.2 (2004): 91-110. - pixels (