JunoCamRawImageProcessing/JitterCorrection.py at master · junoWutopia/JunoCamRawImageProcessing · 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
146
147
148
149
150
151
import numpy as np
import json
from PIL import Image
import cv2  # only needed in detect_sharp_edges

from Util import *

BLUE_JUPITER_THRESHOLD = 35
GREEN_JUPITER_THRESHOLD = 75
RED_JUPITER_THRESHOLD = 75

LAPLACIAN_EDGE_DETECTION_THRESHOLD = 0.1


def get_raw_image_mask(raw_image_array):
    """
    :param raw_image_array: ndarray(k*384, 1648), array of image data from the raw image
    :return: ndarray of uint8, mask of where in the picture jupiter actually is for alignment
    """
    s1, s2 = raw_image_array.shape
    mask = np.zeros((s1, s2), dtype=np.uint8)
    for k in range(0, s1, 384):
        mask[k:k + 128] = (raw_image_array[k:k + 128] > BLUE_JUPITER_THRESHOLD)
        mask[k + 128:k + 256] = (raw_image_array[k + 128:k + 256] > GREEN_JUPITER_THRESHOLD)
        mask[k + 256:k + 384] = (raw_image_array[k + 256:k + 384] > RED_JUPITER_THRESHOLD)
    return mask


def detect_valid_edge_pixels(stripe):
    """
    :param stripe: ndarray(128, 1648), stripe mask
    :return: ndarray(1648) of int, edge indices
             ndarray(1648) of int, mask of which pixels can be used for edge alignment (no noise clear edge)
             ndarray(1648) of int, direction of edge, 1 for edge in up-direction, -1 for edge in down-direction
    """
    stripe = stripe.astype(np.int8)
    y_coords = np.arange(128)
    # detecting the edge
    edge = np.argmax(np.abs(stripe[1:, :]-stripe[:-1, :]), axis=0)
    edge_dir = (stripe[1:, :]-stripe[:-1, :])[edge, np.arange(STRIPE_LENGTH)]

    # finding the indices where values above edge and below edge are all the same by first filling up the rest
    # and the using np.all to detect if the not-filled up part is uniform
    valid_up = np.where(y_coords[:, None] <= edge[None, :], stripe, stripe[None, 0, :])
    valid_down = np.where(y_coords[:, None] > edge[None, :], stripe, stripe[None, -1, :])

    # valid are indices where there are different pixels at the top and bottom
    valid = stripe[None, 0, :] != stripe[None, -1, :]
    valid_up = np.all(valid_up == valid_up[None, 0, :], axis=0)
    valid_down = np.all(valid_down == valid_down[None, 0, :], axis=0)

    return edge, valid_up * valid_down * valid, edge_dir


def detect_sharp_edges(edge, image_stripe, color):
    """
    :param edge: ndarray(1648) of int, edge indices as retuned from detect_valid_edge_pixels
    :param image_stripe: ndarray(128, 1648), stripe of the raw image
    :param color: int (0-2), color of stripe introduced for different edge thresholding for the different colors
                            but not needed yet
    :return: ndarray(1648) of bool, mask on x-dimension similar to second return from detect_valid_edge_pixels
    """
    edge_sharpness = cv2.Laplacian(image_stripe, cv2.CV_32F, ksize=7) / (255 * 49)
    edge_sharpness = cv2.blur(np.abs(edge_sharpness), ksize=(7, 7))

    return edge_sharpness[edge, np.arange(STRIPE_LENGTH)] > LAPLACIAN_EDGE_DETECTION_THRESHOLD


def detect_mask_offset(raw_image_mask, jupiter_mask, raw_image):
    """
    :param raw_image_mask: ndarray of uint8, mask retuned by get_jupiter_mask on the image
    :param jupiter_mask: ndarray of uint8, jupiter mask constructed by position and orientation of spaceprobe
    :return: float, start time offset in seconds for correcting jitter
    """
    s1, s2 = raw_image_mask.shape
    assert jupiter_mask.shape == (s1, s2)

    offset_sum = 0
    offset_count = 0
    for color, k in enumerate(range(0, s1, 128)):
        raw_image_mask_stripe = raw_image_mask[k:k + 128]
        jupiter_mask_stripe = jupiter_mask[k:k + 128]

        jupiter_mask_edge, jupiter_mask_valid, jupiter_mask_egde_dir = detect_valid_edge_pixels(jupiter_mask_stripe)
        raw_image_mask_edge, raw_image_mask_valid, raw_image_egde_dir = detect_valid_edge_pixels(raw_image_mask_stripe)
        sharp_valid = detect_sharp_edges(raw_image_mask_edge, raw_image[k:k + 128], color % 3)

        valid = jupiter_mask_valid * raw_image_mask_valid * (jupiter_mask_egde_dir == raw_image_egde_dir) * sharp_valid
        offset_sum += np.sum((jupiter_mask_edge-raw_image_mask_edge)*valid)
        offset_count += np.sum(valid)

    return offset_sum, offset_count


def correct_image_start_time_and_frame_delay(raw_image_array, raw_image_time_stamp, frame_delay):
    time_offset = STANDARD_IMAGE_START_TIME_OFFSET

    s1, s2 = raw_image_array.shape
    raw_image_mask = get_raw_image_mask(raw_image_array)
    direction_array = np.concatenate(CAMERA_STRIPE_VECTORS, axis=0)

    for i in range(2):
        pixel_offset = 0
        pixel_offset_count = 0
        frame_wise_offsets = []

        for k in range(s1//384):
            cam_pos, cam_orient = get_junocam_jupiter_rel_pos_orient(raw_image_time_stamp,
                                                                     add_seconds=time_offset + k * frame_delay)
            direction_array_new = direction_array.dot(cam_orient)
            _, jupiter_mask = project_onto_jupiter_surf(cam_pos, direction_array_new)
            offset_sum, offset_count = detect_mask_offset(raw_image_mask[k * 384:(k+1) * 384],
                                                          jupiter_mask,
                                                          raw_image_array[k * 384:(k+1) * 384])
            pixel_offset += offset_sum / max(offset_count, 1)
            pixel_offset_count += (offset_count > 0)
            if offset_count != 0:
                frame_wise_offsets.append((k, offset_sum/offset_count))
        # 2rpm = 12deg/sec, approx 5deg per stripe and 128px per stripe
        time_offset += 5/(12*128)*pixel_offset/pixel_offset_count

        # frame delay is changed by mean change in offset per stripe
        tmp = [5/(12*128)*(y[1]-x[1])/(y[0]-x[0]) for x, y in zip(frame_wise_offsets[:-1], frame_wise_offsets[1:])]
        frame_delay += sum(tmp) / len(tmp)
    return time_offset, frame_delay


def correct_image_start_time(raw_image_array, raw_image_time_stamp, frame_delay):
    time_offset = STANDARD_IMAGE_START_TIME_OFFSET

    s1, s2 = raw_image_array.shape
    raw_image_mask = get_raw_image_mask(raw_image_array)
    direction_array = np.concatenate(CAMERA_STRIPE_VECTORS, axis=0)

    for i in range(2):
        pixel_offset = 0
        pixel_offset_count = 0

        for k in range(s1//384):
            cam_pos, cam_orient = get_junocam_jupiter_rel_pos_orient(raw_image_time_stamp,
                                                                     add_seconds=time_offset + k * frame_delay)
            direction_array_new = direction_array.dot(cam_orient)
            _, jupiter_mask = project_onto_jupiter_surf(cam_pos, direction_array_new)
            offset_sum, offset_count = detect_mask_offset(raw_image_mask[k * 384:(k+1) * 384],
                                                          jupiter_mask,
                                                          raw_image_array[k * 384:(k+1) * 384])
            pixel_offset += offset_sum / max(offset_count, 1)
            pixel_offset_count += (offset_count > 0)
        # 2rpm = 12deg/sec, approx 5deg per stripe and 128px per stripe
        time_offset += 5/(12*128)*pixel_offset/pixel_offset_count
    return time_offset