-4
私はこれらの2つの機能を終了しようとしていますが、誰かが私にそれを行う方法を示すことができる場合はどこから始めたらいいのですか？反射境界と相互相関2Dをどのように定義しますか？
def padReflectBorder(image, N):
"""This function pads the borders of the input image by reflecting the
image across the boundaries.
N is the number of rows or columns that should be added at each border;
i.e., the output size should have 2N more rows and 2N more columns than
the input image.
The values in the input image should be copied to fill the middle of the
larger array, and the borders should be filled by reflecting the array
contents as described in the documentation for cv2.copyMakeBorder().
This function should be equivalent to the library call:
cv2.copyMakeBorder(image, N, N, N, N, borderType=cv2.BORDER_REFLECT_101)
Note: BORDER_REFLECT_101 means that the values in the image array are
reflected across the border. Ex. gfedcb|abcdefgh|gfedcba
NOTE: You MAY NOT use any calls to numpy or opencv library functions, but
you MAY use array broadcasting and "advanced" numpy indexing
techniques for this function.
Parameters
----------
image : numpy.ndarray(dtype=np.uint8)
A grayscale image represented in a numpy array.
N : int
An integer strictly greater than zero and less than the smallest
dimension of the input image representing the number of padding pixels
to add at each border.
Returns
-------
numpy.ndarray(dtype=np.uint8)
A copy of the input array with 2N additional rows and columns filled
with the values of the input image reflected over the borders.
"""
def crossCorrelation2D(image, kernel):
"""This function uses native Python code & loops to compute and return the
valid region of the cross correlation of an input kernel applied to each
pixel of the input array.
NOTE: Lectures 2-05, 2-06, and 2-07 address this concept.
Recall that for an image F and kernel h, cross correlation is defined as:
G(i,j) = sum_u=-k..k sum_v=-k..k h[u,v] F[i+u,j+v]
For N = kernel.shape[0] // 2, this function should be equivalent to:
cv2.filter2D(image, cv2.CV_64F, kernel)[N:-N, N:-N]
See http://docs.opencv.org/2.4/modules/imgproc/doc/filtering.html#filter2d
for details.
Your code must operate on each pixel of the image and kernel individually
for each step of the computation. (We know this is inefficient, but we want
to make sure that you understand what is really happening within the more
efficient library functions that are available.)
NOTE: You MAY NOT use any numpy, scipy, or opencv library functions,
broadcasting rules, or "advanced" numpy indexing techniques, nor may
you use the operator functions or convolution functions listed in the
note at the top. You MUST manually loop through the image at each
pixel. (Yes, we know this is slow and inefficient.)
NOTE: You MAY assume that kernel will always be a square array with an odd
number of elements.
Parameters
----------
image : numpy.ndarray(dtype=np.uint8)
A grayscale image represented in a numpy array.
kernel : numpy.ndarray
A kernel represented in a numpy array of size (k, k) where k is an odd
number strictly greater than zero.
Returns
-------
output : numpy.ndarray(dtype=np.float64)
The output image. The size of the output array should be smaller than
the original image size by k-1 rows and k-1 columns, where k is the
size of the kernel.
"""