Merge branch 'github/fork/gsn9/mutualInformation' into 'subpixelapi'

from math import floor

from autocnet.transformation.roi import Roi

import numpy as np

from scipy.ndimage.measurements import center_of_mass

from skimage.transform import AffineTransform

def mutual_information(t1, t2, **kwargs):

def mutual_information(reference_arr, moving_arr, affine=AffineTransform(), **kwargs):

    """

    Computes the correlation coefficient between two images using a histogram

    comparison (Mutual information for joint histograms). The corr_map coefficient

    Parameters

    ----------

    t1 : ndarray

    reference_arr : ndarray

                    First image to use in the histogram comparison

    t2 : ndarray

    moving_arr : ndarray

                   Second image to use in the histogram comparison

    Returns

    -------

    numpy.histogram2d : for the kwargs that can be passed to the comparison

    """

    if np.isnan(t1).any() or np.isnan(t2).any():

    if np.isnan(reference_arr.data).any() or np.isnan(moving_arr.data).any():

        print('Unable to process due to NaN values in the input data')

        return

    if t1.shape != t2.shape:

    if reference_arr.shape != moving_arr.shape:

        print('Unable compute MI. Image sizes are not identical.')

        return

    hgram, x_edges, y_edges = np.histogram2d(t1.ravel(),t2.ravel(), **kwargs)

    hgram, x_edges, y_edges = np.histogram2d(reference_arr.ravel(),moving_arr.ravel(), **kwargs)

    # Convert bins counts to probability values

    pxy = hgram / float(np.sum(hgram))

    nzs = pxy > 0 # Only non-zero pxy values contribute to the sum

    return np.sum(pxy[nzs] * np.log(pxy[nzs] / px_py[nzs]))

def mutual_information_match(d_template, s_image, subpixel_size=3,

# TODO 

# need's to take in a ROI and not ndarray's

# and use one clip (to pass arr later on?)

def mutual_information_match(moving_roi, reference_roi, subpixel_size=3,

                             func=None, **kwargs):

    """

    Applys the mutual information matcher function over a search image using a

    Parameters

    ----------

    d_template : ndarray

    moving_roi : roi 

                 The input search template used to 'query' the destination

                 image

    s_image : ndarray

    reference_roi : roi

              The image or sub-image to be searched

    subpixel_size : int

    Returns

    -------

    x : float

        The x offset

    y : float

        The y offset

    new_affine :AffineTransform

                The affine transformation

    max_corr : float

               The strength of the correlation in the range [0, 4].

               Map of corrilation coefficients when comparing the template to

               locations within the search area

    """

    reference_template = reference_roi.clip()

    moving_image = moving_roi.clip()

    if func == None:

        func = mutual_information

    image_size = s_image.shape

    template_size = d_template.shape

    image_size = moving_image.shape

    template_size = reference_template.shape

    y_diff = image_size[0] - template_size[0]

    x_diff = image_size[1] - template_size[1]

    max_corr = -np.inf

    corr_map = np.zeros((y_diff+1, x_diff+1))

    max_i = -1  # y

    max_j = -1  # x

    for i in range(y_diff+1):

        for j in range(x_diff+1):

            sub_image = s_image[i:i+template_size[1],  # y

            sub_image = moving_image[i:i+template_size[1],  # y

                                j:j+template_size[0]]  # x

            corr = func(sub_image, d_template, **kwargs)

            corr = func(sub_image, reference_template, **kwargs)

            if corr > max_corr:

                max_corr = corr

                max_i = i

                max_j = j

            corr_map[i, j] = corr

    y, x = np.unravel_index(np.argmax(corr_map, axis=None), corr_map.shape)

    upper = int(2 + floor(subpixel_size / 2))

    lower = upper - 1

    # x, y are the location of the upper left hand corner of the template in the image

    area = corr_map[y-lower:y+upper,

                    x-lower:x+upper]

    # Compute the y, x shift (subpixel) using scipys center_of_mass function

    cmass  = center_of_mass(area)

    if area.shape != (subpixel_size + 2, subpixel_size + 2):

        print("Max correlation is too close to the boundary.")

        return None, None, 0, None

        return  None, 0, None

    subpixel_y_shift = subpixel_size - 1 - cmass[0]

    subpixel_x_shift = subpixel_size - 1 - cmass[1]

    y = abs(y - (corr_map.shape[1])/2)

    x = abs(x - (corr_map.shape[0])/2)

    y += subpixel_y_shift

    x += subpixel_x_shift

    # Compute the idealized shift (image center)

    y -= (s_image.shape[0] / 2) - (d_template.shape[0] / 2)

    x -= (s_image.shape[1] / 2) - (d_template.shape[1] / 2)

    return float(x), float(y), float(max_corr), corr_map

    new_affine = AffineTransform(translation=(-x, -y))

    return new_affine, np.max(max_corr), corr_map

 No newline at end of file

import sys

import unittest

from unittest.mock import patch

from autocnet.transformation.roi import Roi

import pytest

import numpy as np

def test_bad_mi():

    test_image1 = np.array([[i for i in range(50)] for j in range(50)])

    test_image2 = np.ones((50, 50))

    corrilation = mutual_information.mutual_information(test_image1, test_image2)

    assert corrilation == pytest.approx(0)

    s_image = np.ones((100, 100))

    s_image[25:75, 25:75] = d_template

    reference_roi  = Roi(d_template, 25, 25, 25, 25, ndv=22222222)

    moving_roi = Roi(s_image, 50, 50, 50, 50, ndv=22222222)

    x_offset, y_offset, max_corr, corr_map = mutual_information.mutual_information_match(d_template, s_image, bins=20)

    assert x_offset == 0.01711861257171421

    assert y_offset == 0.0

    affine, max_corr, corr_map = mutual_information.mutual_information_match(moving_roi, reference_roi, bins=20)

    assert affine.params[0][2] == -0.5171186125717124

    assert affine.params[1][2] == -0.5

    assert max_corr == 2.9755967600033015

    assert corr_map.shape == (51, 51)

    assert np.min(corr_map) >= 0.0

    lons, lats = isis.image_to_ground(reference_image.file_name, x_coords, y_coords, allowoutside=True)

    xs, ys = isis.ground_to_image(moving_image.file_name, lons, lats, allowoutside=True)

    # Check for any coords that do not project between images

    base_gcps = []

    dst_gcps = []