UCI-ALLEN-BrainRepositoryCode/stitcher.py at main · andythai/UCI-ALLEN-BrainRepositoryCode · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
"""
Code provided from Allen Institute
"""


import logging
import operator as op
from collections import defaultdict
from six.moves import reduce

import numpy as np

class Stitcher(object):


    def __init__(self, image_dimensions, tiles, channels):

        logging.info('image_dimensions: {0}'.format(image_dimensions))
        self.image_dimensions = image_dimensions


        self.tiles = tiles
        self.channels = channels


    def run(self, cb=np.array):

        slice_image, stitched_indicator = initialize_images(self.image_dimensions, len(self.channels))
        missing_tiles = {}

        for tile in self.tiles:

            if tile.is_missing:

                missing_tiles[tile.index] = tile.get_missing_path()
                tile.initialize_image()

            else:
                tile.trim_self()

            self.stitch(slice_image, stitched_indicator, tile, cb)

        return slice_image, missing_tiles


    def stitch(self, slice_image, stitched_indicator, tile, cb=np.array):

        region = tile.get_image_region()

        current_region = slice_image[region[0], region[1], region[2]]
        indicator_region = stitched_indicator[region[0], region[1], region[2]]

        stup = (tile.size['row'], tile.size['column'])
        blend = get_blend(indicator_region, stup, cb)
        blend = make_blended_tile(blend, tile.image, current_region)
        #logging.info('obtained blend')

        slice_image[region[0], region[1], region[2]] = blend
        stitched_indicator[region[0], region[1], region[2]] = 1
        #logging.info('updated image region with tile data')


def initialize_image(dimensions, nchannels, dtype, order='C'):
    return np.zeros((dimensions['row'], dimensions['column'], nchannels), dtype=dtype, order=order)


def initialize_images(dimensions, nchannels):
    return initialize_image(dimensions, nchannels, np.uint16), initialize_image(dimensions, nchannels, np.int8)


def make_blended_tile(blend, tile, current_region):
    return np.multiply((1 - blend), tile) + np.multiply(blend, current_region)


def get_indicator_bound_point(indicator, lg, axis):
    '''Finds the index of first change in a binary mask
    along a specified axis in a specified direction
    '''

    delta = np.diff(indicator, axis=axis)
    points = np.where(lg(delta, 0))
    del delta

    points = np.unique(points[axis])
    size = indicator.shape[axis]
    points = points[lg(points, size / 2.0)]

    if len(points) > 0:
        return points[-1]
    return None


def blend_component_from_point(point, mesh, lg):
    '''Obtains a normalized component of the blend, which describes depth of
    overlap along a specified axis in a specified direction
    '''

    # this has the effect that the shallowest part of the blend
    # is always 0 - symmetric with the deepest after normalization.
    blend = point - mesh + 1
    blend[lg(blend, 0)] = 0

    blend = np.fabs(blend)
    mx = np.amax(blend)
    mx = mx if mx > 0.0 else 1.0

    return blend / mx


def get_blend_component(indicator, lg, axis, meshes):
    '''
    '''

    point = get_indicator_bound_point(indicator, lg, axis)
    if point is None:
        return []

    return [blend_component_from_point(point, meshes[axis], lg)]


def get_overall_blend(indicator, meshes):
    '''
    '''

    blends = []

    for lg in (op.lt, op.gt):
        for axis in (0, 1):
            blends.extend(get_blend_component(indicator, lg, axis, meshes))

    if len(blends) == 0:
        return np.zeros_like(indicator)
    return reduce(np.maximum, blends)


def get_blend(indicator_region, stup, cb=np.array):
    '''
    '''

    meshes = np.meshgrid(*map(np.arange, stup), indexing='ij')
    blend = get_overall_blend(indicator_region, meshes)

    return cb(np.multiply(blend, indicator_region))