Module integralimage

Convert an image of type uint8 into an image integral of type double.

Input:

img: a 2D numpy.array of type uint8 iimg: a 2D numpy.array of type double of the same size to hold the data

Output:

data filled into img2, where iimg[x,y] = int( img[x,y], i = 0..x, j = 0..y )

Convert an image of type uint8 into an image integral of type uint32.

Input:

img: a 2D numpy.array of type uint8 iimg: a 2D numpy.array of type uint32 of the same size to hold the data

Output:

data filled into img2, where iimg[x,y] = int( img[x,y], i = 0..x, j = 0..y )

Convert an image of type uint8 into an image integral of type int32.

Input:

img: a 2D numpy.array of type uint8 iimg: a 2D numpy.array of type int32 of the same size to hold the data

Output:

data filled into img2, where iimg[x,y] = int( img[x,y], i = 0..x, j = 0..y )

Given a list of image patches, integrate then vectorize them all.

Input:

input_data: a numpy.array of shape (N,d,d) type uint8

Output:

output_data: the vectorized-integrated version, a numpy.array of shape (N,d*d) type float64

Similar to stats.statslist(), but integrate each image of type uint8 first, then vectorize, then do the stats.

Input:

input_data: a array 2D images of type uint8 (numpy.array)

Output:

w, mean, cov: total weight, mean vector and covariance matrix

Given a J-class ClassificationDataset of image patches. The patches are either in 'uint8' raw form (vi = False), or in 'double' vectorintegrated form (vi = True)

Input:

classification_dataset: a ClassificationDataset vi: input data are vectorintegrated?

Output:

stats2: a Stats2 holding the obtained statistics