def export_vp_stack(self, type='png', downscale=None):
# print status
self.sta.status_msg('Write viewpoint image stack',
self.cfg.params[self.cfg.opt_prnt])
self.sta.progress(None, self.cfg.params[self.cfg.opt_prnt])
# downscale image
downscale = True if downscale is None else downscale
views_stacked_img = misc.img_resize(self.views_stacked_img.copy(), 1 / self._M) \
if downscale else self.views_stacked_img.copy()
# normalization
p_lo = np.percentile(misc.rgb2gray(self.central_view), 0.05)
p_hi = np.percentile(misc.rgb2gray(self.central_view), 99.995)
views_stacked_img = misc.Normalizer(views_stacked_img,
min=p_lo,
max=p_hi).uint8_norm()
# export all viewpoints in single image
views_stacked_path = os.path.join(
self.cfg.exp_path, 'views_stacked_img_' + str(self._M) + 'px')
misc.save_img_file(views_stacked_img,
file_path=views_stacked_path,
file_type=type)
self.sta.progress(100, self.cfg.params[self.cfg.opt_prnt])
return True
def crop_imgs(folder, coords_lists):
img_tiles, files = list(), list()
exts = ('tif', 'tiff', 'png')
misc.mkdir_p(os.path.join(folder, 'auto-crop'))
files = [
f for f in os.listdir(folder)
if f.endswith(exts) and not f.startswith('.')
]
files.sort()
for i, file in enumerate(files):
coords_nested = coords_lists[i]
for j, coords in enumerate(coords_nested):
if coords[0] != 0 and coords[1] != 0:
cy, cx, h, w = coords
img = misc.load_img_file(os.path.join(folder, file))
tile = img[cy - h // 2:cy + h // 2, cx - w // 2:cx + w // 2,
...]
img_tiles.append(tile)
fn, ext = os.path.splitext(file)
misc.save_img_file(tile,
os.path.join(folder, 'auto-crop',
fn + '_crop' + str(j)),
'png',
tag=True)
return img_tiles, files
def export_thumbnail(self, type='tiff'):
# export central viewpoint as thumbnail image
fp = os.path.join(self.cfg.exp_path, 'thumbnail')
misc.save_img_file(self.central_view, file_path=fp, file_type=type)
return True
def export_viewpoints(self, type='tiff'):
# print status
self.sta.status_msg('Write viewpoint images', self.cfg.params[self.cfg.opt_prnt])
self.sta.progress(None, self.cfg.params[self.cfg.opt_prnt])
ptc_leng = self.cfg.params[self.cfg.ptc_leng]
# create folder
folderpath = os.path.join(self.cfg.exp_path, 'viewpoints_'+str(ptc_leng)+'px')
misc.mkdir_p(folderpath)
# normalize image array to 16-bit unsigned integer
vp_img_arr = Normalizer(self.vp_img_arr).uint16_norm()
# export viewpoint images as image files
for j in range(ptc_leng):
for i in range(ptc_leng):
misc.save_img_file(vp_img_arr[j, i], os.path.join(folderpath, str(j) + '_' + str(i)), file_type=type)
# print status
percentage = (((j*self._M+i+1)/self._M**2)*100)
self.sta.progress(percentage, self.cfg.params[self.cfg.opt_prnt])
return True
def desaturate_clipped(img_arr, gains=None):
b, r, g1, g2 = gains if gains is not None else [1., 1., 1., 1.]
awb_list = [r, g1, b]
comp_arr = np.zeros(img_arr.shape, dtype=img_arr.dtype)
mask = np.zeros(img_arr.shape[:2])
alpha = np.amin(img_arr / np.amax(img_arr, axis=2)[..., np.newaxis],
axis=2)
thresh = .99
for i in range(img_arr.shape[2]):
comp_arr[..., i] = (1 - alpha) * (img_arr[..., i]) + alpha * (
img_arr[..., i] / awb_list[i])
# determine indices of other two channels
ch1, ch2 = [u for u, v in enumerate(awb_list) if v != awb_list[i]]
# only consider pixels in mask which have sufficient intensity distance
mask[(img_arr[..., i] > img_arr[..., i].max() * thresh)
& (img_arr[..., ch1] > img_arr[..., i].max() * thresh) &
(img_arr[..., ch2] > img_arr[..., i].max() * thresh)] = 1
import os
misc.save_img_file(mask, os.getcwd() + 'mask.png')
img_arr[mask > 0] = comp_arr[mask > 0]
return img_arr
def export_refo_stack(self, type=None):
refo_stack = Normalizer(np.asarray(self._refo_stack)).uint16_norm()
# create folder
string = ''
if self.cfg.params[self.cfg.opt_refo] == 2 or self.cfg.params[
self.cfg.opt_refi]:
string = 'subpixel_'
elif self.cfg.params[self.cfg.opt_refo] == 1 or self.cfg.params[
self.cfg.opt_view]:
string = ''
folder_path = os.path.join(self.cfg.exp_path,
'refo_' + string + str(self._M) + 'px')
misc.mkdir_p(folder_path)
for i, refo_img in enumerate(refo_stack):
# get depth plane number for filename
a = range(*self.cfg.params[self.cfg.ran_refo])[i]
# account for sub-pixel precise depth value
a = round(float(a) /
self._M, 2) if self.cfg.params[self.cfg.opt_refi] else a
# write image file
misc.save_img_file(refo_img,
os.path.join(folder_path, str(a)),
file_type=type)
return True
def export_refo_stack(refo_stack, cfg, type='tiff'):
refo_stack = misc.uint16_norm(refo_stack)
patch_len = cfg.params[cfg.ptc_leng]
# create folder
string = ''
if cfg.params[cfg.opt_refo] == 2 or cfg.params[cfg.opt_refi]:
string = 'subpixel_'
elif cfg.params[cfg.opt_refo] == 1 or cfg.params[cfg.opt_view]:
string = ''
folder_path = os.path.join(cfg.params[cfg.lfp_path].split('.')[0], 'refo_' + string + str(patch_len) + 'px')
misc.mkdir_p(folder_path)
for i, refo_img in enumerate(refo_stack):
# get depth plane number for filename
a = range(*cfg.params[cfg.ran_refo])[i]
# account for sub-pixel precise depth value
a = round(float(a)/patch_len, 2) if cfg.params[cfg.opt_refi] else a
# write image file
misc.save_img_file(refo_img, os.path.join(folder_path, str(a)), type=type)
return True
def export_thumbnail(vp_img_arr, cfg, type='tiff'):
# export central viewpoint as thumbnail image
misc.save_img_file(vp_img_arr[cfg.params[cfg.ptc_leng]//2+1, cfg.params[cfg.ptc_leng]//2+1],
os.path.join(cfg.params[cfg.lfp_path].split('.')[0], 'thumbnail'), type=type)
return True
def scheimpflug_from_stack(self):
a_start, a_stop = (0, len(self.refo_stack))
if len(self.refo_stack[0].shape) == 3:
m, n, p = self.refo_stack[0].shape
else:
m, n, p = (self.refo_stack[0].shape[0],
self.refo_stack[0].shape[1], 1)
scheimpflug_img = np.zeros([m, n, p])
# map generation
a_x = np.linspace(
0, a_stop - a_start, n + 2,
dtype='int')[1:-1] # flooring via integer type while excluding
a_y = np.linspace(0, a_stop - a_start, m + 2, dtype='int')[1:-1]
a_map_x = np.outer(np.ones(m, dtype='int'), a_x)
a_map_y = np.outer(a_y, np.ones(n, dtype='int'))
# vertical orientation (default)
a_map = a_map_y
# horizontal orientation
if self.cfg.params[self.cfg.opt_pflu] == c.PFLU_VALS[1]:
a_map = a_map_x
# diagonal orientation
elif self.cfg.params[self.cfg.opt_pflu] == (c.PFLU_VALS[2]
or c.PFLU_VALS[3]):
# swap refocusing directions if option set
if self.cfg.params[self.cfg.opt_pflu] == c.PFLU_VALS[3]:
a_map_x, a_map_y = a_map_x[::-1], a_map_y[::-1]
a_map = np.mean(np.stack([a_map_x, a_map_y]), dtype='int', axis=0)
for y in range(m):
for x in range(n):
scheimpflug_img[y, x] = self.refo_stack[a_map[y, x]][y, x]
# check interrupt status
if self.sta.interrupt:
return False
# print status
percentage = (((y * n + x + 1) / (m * n)) * 100)
self.sta.progress(percentage,
self.cfg.params[self.cfg.opt_prnt])
# write image file to hard drive
a_ran = self.cfg.params[self.cfg.ran_refo]
fn = 'scheimpflug_' + str(a_ran[0]) + '_' + str(
a_ran[-1]) + '_' + self.cfg.params[self.cfg.opt_pflu] + '.png'
misc.save_img_file(
misc.Normalizer(scheimpflug_img).uint16_norm(),
os.path.join(self.fp, fn))
return True
def export_thumbnail(self, type='tiff'):
thumb = misc.Normalizer(self.central_view.copy()).type_norm()
# export central viewpoint as thumbnail image
fp = os.path.join(self.cfg.exp_path, 'thumbnail')
misc.save_img_file(thumb,
file_path=fp,
file_type=type,
tag=self.cfg.params[self.cfg.opt_dbug])
return True
def rectify_candidates_bayer(self, bay_img=None, n=2, sig_lev=4):
# status message
self.sta.status_msg('Hot pixel detection',
self.cfg.params[self.cfg.opt_prnt])
bay_img = self._bay_img.copy() if bay_img is None else bay_img
cum_img = np.zeros(bay_img[0::2, 0::2].shape)
for c in range(4):
i, j = c // 2, c % 2
# progress update
percent = c / 4
self.sta.progress(percent * 100,
self.cfg.params[self.cfg.opt_prnt])
# deduct median filtered image
med_img = medfilt(bay_img[i::2, j::2].copy(), kernel_size=(3, 3))
m_img = bay_img[i::2, j::2].copy() / bay_img[
i::2, j::2].max() - med_img.copy() / med_img.max()
new_img = self.rectify_candidates_channel(
channel=bay_img[i::2, j::2].copy(),
ref_img=m_img,
med_img=med_img,
n=n,
sig_lev=sig_lev + 2)
diff_img = bay_img[i::2, j::2].copy() - new_img
cum_img[diff_img != 0] = 1
bay_img[i::2, j::2] = new_img
# check interrupt status
if self.sta.interrupt:
return False
# export hot-pixel map
if self.cfg.params[self.cfg.opt_dbug]:
hotp_num = len(cum_img[cum_img != 0])
misc.save_img_file(cum_img,
file_path=os.path.join(
self.cfg.exp_path,
'hot-pixel_map_' + str(hotp_num) + '.png'))
# progress update
self.sta.progress(100, self.cfg.params[self.cfg.opt_prnt])
self._bay_img = bay_img
return True
def write_centroids_img(self, fn='default_filename.png'):
# draw MICs in binary image and save image file for debug purposes
if self.cfg.params[self.cfg.opt_dbug]:
plot_img = self.draw_centroids_img()
save_img_file(
plot_img,
os.path.join(
os.path.splitext(self.cfg.params[self.cfg.lfp_path])[0],
fn))
# plot_centroids(img, centroids)
return True
def export_vp_stack(self, type='tiff'):
# print status
self.sta.status_msg('Write viewpoint image stack', self.cfg.params[self.cfg.opt_prnt])
self.sta.progress(None, self.cfg.params[self.cfg.opt_prnt])
# export all viewpoints in single image
views_stacked_path = os.path.join(self.cfg.exp_path, 'views_stacked_img_'+str(self._M)+'px')
misc.save_img_file(misc.img_resize(self.views_stacked_img, 1/self._M),
file_path=views_stacked_path, file_type=type)
self.sta.progress(100, self.cfg.params[self.cfg.opt_prnt])
return True
def _write_lfp_align(self):
# convert to 16bit unsigned integer
self._lfp_out = misc.Normalizer(self._lfp_out).uint16_norm()
# create output data folder
misc.mkdir_p(self.cfg.exp_path, self.cfg.params[self.cfg.opt_prnt])
# write aligned light field as pickle file to avoid recalculation
with open(os.path.join(self.cfg.exp_path, 'lfp_img_align.pkl'), 'wb') as f:
pickle.dump(self._lfp_out, f)
if self.cfg.params[self.cfg.opt_dbug]:
misc.save_img_file(self._lfp_out, os.path.join(self.cfg.exp_path, 'lfp_img_align.tiff'))
def save_refo_slice(self, a, refo_img, folder_path=None, file_type=None, string=None):
string = 'subpixel_' if self.cfg.params[self.cfg.opt_refi] and string is None else string
if folder_path is None:
folder_path = os.path.join(self.cfg.exp_path, 'refo_' + string + str(self._M) + 'px')
misc.mkdir_p(folder_path)
# account for sub-pixel precise depth value
a = round(float(a) / self._M, 2) if self.cfg.params[self.cfg.opt_refi] else a
# write image file
misc.save_img_file(refo_img, os.path.join(folder_path, str(a)), file_type=file_type)
return True
def write_centroids_img(self, fn='centroids_img.png'):
# status message handling
self.sta.status_msg(msg='Save centroids image', opt=self.cfg.params[self.cfg.opt_prnt])
self.sta.progress(None, self.cfg.params[self.cfg.opt_prnt])
# draw MICs in binary image and save image file for debug purposes
plot_img = self.draw_centroids_img()
save_img_file(plot_img, os.path.join(os.path.splitext(self.cfg.params[self.cfg.lfp_path])[0], fn))
# self.plot_centroids()
# status message handling
self.sta.progress(100, self.cfg.params[self.cfg.opt_prnt])
return True
def export_viewpoints(vp_img_arr, cfg, type='tiff'):
ptc_leng = cfg.params[cfg.ptc_leng]
# create folder
folderpath = os.path.join(cfg.params[cfg.lfp_path].split('.')[0], 'viewpoints_' + str(ptc_leng) + 'px')
misc.mkdir_p(folderpath)
# normalize image array to 16-bit unsigned integer
vp_img_arr = misc.uint16_norm(vp_img_arr)
# export viewpoint images as image files
for j in range(ptc_leng):
for i in range(ptc_leng):
misc.save_img_file(vp_img_arr[j, i], os.path.join(folderpath, str(j) + '_' + str(i)), type=type)
return True
def save_refo_slice(self,
a,
refo_img,
folder_path=None,
file_type=None,
string=None):
string = 'subpixel_' if self.cfg.params[
self.cfg.opt_refi] and string is None else string
if folder_path is None:
folder_path = os.path.join(self.cfg.exp_path,
'refo_' + string + str(self._M) + 'px')
misc.mkdir_p(folder_path)
# write image file
misc.save_img_file(refo_img,
os.path.join(folder_path, str(a)),
file_type=file_type)
return True
def decode_lytro_file(self):
# Lytro type decoding
with open(self._lfp_path, mode='rb') as file:
# LFC and raw type decoding
obj = LfpDecoder(file, self.cfg, self.sta, lfp_path=self._lfp_path)
obj.main()
self._lfp_img = obj.bay_img
self._json_dict = obj.json_dict
del obj
# save bayer image as file (if not already present)
if not os.path.exists(self.fp) and not self.sta.interrupt:
self.sta.status_msg(msg='Save raw image',
opt=self.cfg.params[self.cfg.opt_prnt])
self.sta.progress(None, self.cfg.params[self.cfg.opt_prnt])
misc.save_img_file(misc.Normalizer(
self._lfp_img).uint16_norm(),
self.fp,
file_type='tiff')
self.sta.progress(100, self.cfg.params[self.cfg.opt_prnt])
return True
def ver_hex_bulge(self, img):
# variable init
img = np.divide(img.astype('float'), img.max(), out=np.zeros_like(img), where=img != 0)
m, n, ch = img.shape if len(img.shape) == 3 else img.shape + (1,)
# j for shifted and i represents truth (unshifted)
i, j = [1, 0] if self.hex_odd else [0, 1]
# vertically interpolate pixels to get reference
#phi = np.zeros((img[i::2, ...].shape[0], img.shape[1], img.shape[2]))
#for p in range(ch):
# fun = interp2d(range(img[i::2, ...].shape[1]), range(img[i::2, ...].shape[0]), img[i::2, :, p], kind=self.method)
# phi[..., p] = fun(np.arange(img.shape[1]), np.linspace(0, img[i::2, ...].shape[0]+.5, img[i::2, ...].shape[0]))
# gradient between shifted and unshifted (responsive to hex artifacts)
dif = img[j::2, ...]-img[i::2, ...] if img[i::2, ...].shape[0] == img[j::2, ...].shape[0] else img[j::2, ...]-img[i::2, ...][:-1]
# gradient between unshifted (responsive to real object edges)
edg = img[i::2, ...][:-1]-img[i+2::2, ...]
edg *= 0.75
# absolute values
dif = abs(dif)
edg = abs(edg)
# (don't throw away channel information)
dif = dif.sum(2)
edg = edg.sum(2)
# deduct actual edges
karr = dif - edg if dif.shape[0] == edg.shape[0] else dif[:-1, ...] - edg
karr = np.divide(karr, karr.max(), out=np.zeros_like(karr), where=karr != 0)
karr[karr < 0] = 0
if self.cfg.params[self.cfg.opt_dbug]:
misc.save_img_file(karr, os.path.join(self.cfg.exp_path, 'karr.png'))
# replicate mask to 3 color channels
arr = np.zeros(karr.shape+(3,)) #if karr.shape[0] == dif.shape[0] else np.zeros_like(dif[:-1, ...])
for p in range(ch):
arr[..., p] = karr
# form long col vector as we are only interested in vertical direction
arr_vec = arr.copy().T.ravel()
# remove spikes (fast peaks in vertical direction)
#arr_med = medfilt(arr_vec, 3)
# 1-D low pass to eliminate hi freq noise
#gauss = misc.create_gauss_kernel(10, sig=1)[:, 3]
#arr_gau = np.convolve(arr_med, gauss, 'same')
arr_res = arr_vec.reshape(arr.T.shape).T
#rat_res = rat_gau.reshape(rat.T.shape).T
# threshold values with insufficient difference
mask = np.divide(arr_res, arr_res.max(), out=np.zeros_like(arr_res), where=arr_res != 0)
th = 0.1
mask[mask < th] = 0
mask[mask >= th] = 1
# ignore small regions
mask = self.retain_connected(mask, n=4)
# generate mask
if self.cfg.params[self.cfg.opt_dbug]:
full_mask = np.zeros(img.shape)
if full_mask[j::2, ...].shape[0] == mask.shape[0]:
full_mask[j::2, ...] = mask
else:
full_mask[j::2, ...][:-1] = mask
misc.save_img_file(full_mask, os.path.join(self.cfg.exp_path, 'hex_filter_mask.png'), tag=True)
# generate image indicating which pixels were treated
if self.cfg.params[self.cfg.opt_dbug]:
tes = img.copy()
for p in range(ch):
if tes[j::2, ...].shape[0] == mask.shape[0]:
tes[j::2, :, p][mask[..., p] != 0] = tes.max()
else:
tes[j::2, :, p][:-1][mask[..., p] != 0] = tes.max()
tes_out = np.divide(tes, tes.max(), out=np.zeros_like(tes), where=tes != 0)
misc.save_img_file(tes_out, os.path.join(self.cfg.exp_path, 'ident.png'), tag=True)
idxs = np.array(mask.nonzero())
r = 1
valid_idx = np.where((idxs[0] > r) & (idxs[1] > r) & (img[j::2].shape[0]-idxs[0] > r) & (img.shape[1]-idxs[1] > r))[0]
idxs_vals = list(zip(idxs[0][valid_idx], idxs[1][valid_idx], idxs[2][valid_idx]))
res = img.copy()
arr_res = np.divide(arr_res, arr_res.max(), out=np.zeros_like(arr_res), where=arr_res != 0)
for idx in idxs_vals:
#f = arr_res[idx[0], idx[1], idx[2]]
#new_val = res[j::2, ...][idx[0], idx[1], idx[2]] * (1-f) + phi[idx[0], idx[1], idx[2]] * f #shi[j::2, ...][idx[0], idx[1]]#np.mean(img[i::2, ...][idx[0]:idx[0]+2, idx[1]], axis=0)
new_val = np.mean(img[i::2, ...][idx[0]:idx[0]+2, idx[1], idx[2]], axis=0)
#new_val = np.mean(img[i::2, ...][idx[0]:idx[0]+2, idx[1]-1:idx[1]+2, idx[2]], axis=(0, 1))
res[j::2, ...][idx[0], idx[1], idx[2]] = new_val
if self.cfg.params[self.cfg.opt_dbug]:
misc.save_img_file(res, os.path.join(self.cfg.exp_path, 'hex-corr.png'))
return res
def scheimpflug_from_scratch(self):
patch_len = self.cfg.params[self.cfg.ptc_leng]
a_start, a_stop = self.cfg.params[self.cfg.ran_refo]
img = self.lfp_img.astype('float') / patch_len
m, n, P = self.lfp_img.shape
overlap = int(abs(a_stop) * (patch_len - 1))
#tilt_vec_y = np.linspace(a_start, planes, m).astype('uint')
a_x = np.linspace(a_start, a_stop, n + overlap).astype('int') # flooring values instead of rounding => 1st round, then int()
a_y = np.zeros(m).astype('int')#np.linspace(a_start, planes, m).astype('uint')
a_xy = np.ones([m, n + overlap]).astype('int') * a_x
# initialize matrices for intermediate and plane results
hor_refo = np.zeros([m, n + overlap, P], dtype='float')
ver_refo = np.zeros([m + overlap, n + overlap, P], dtype='float')
fraction_vec = np.zeros([patch_len, P], dtype='float')
# horizontal scheimpflug shift and integration
for y in range(m):
for x in range(n):
# prevent from taking adjacent pixel being beyond image border
if x + patch_len < n:
adj_idx = x + patch_len
else:
adj_idx = x
for p in range(P):
fraction_vec[:, p] = np.linspace(img[y, x, p], img[y, adj_idx, p], patch_len)
# load refocus value at x,y
a = a_xy[y, x]
# consider negative shift
negative_a = int((patch_len - 1) * abs(a)) if a < 0 else 0
# centralize row
row_shift = int((a_y.max() - a)*(patch_len - 1)/2)
newX = int(x + np.mod(x, patch_len) * (a - 1) + negative_a) + row_shift
hor_refo[y, newX:newX + patch_len, :] = hor_refo[y, newX:newX + patch_len, :] + fraction_vec
# check interrupt status
if self.sta.interrupt:
return False
# print progress status
self.sta.progress((y+1)/m*50, self.cfg.params[self.cfg.opt_prnt])
# vertical scheimpflug shift and integration
new_n = n + int(abs(a_x.max()) * (patch_len - 1))
for x in range(new_n):
for y in range(m):
if y + patch_len < m:
adj_idx = y + patch_len
else:
adj_idx = y
frac_vec = np.zeros([patch_len, int(n + abs(a_x.max()) * (patch_len - 1)), P], dtype='float')
for p in range(P):
frac_vec[:, x, p] = np.linspace(hor_refo[y, x, p], hor_refo[adj_idx, x, p], patch_len)
# load refocus value at x,y
a = a_xy[y, x]
# consider negative shift
negative_a = int((patch_len - 1) * abs(a)) if a < 0 else 0
# centralize column
column_shift = int((a_x.max()-a)*(patch_len-1)/2)
# put interpolated vector to refocusing plane
newY = int(y + np.mod(y, patch_len) * (a - 1) + negative_a) + column_shift
ver_refo[newY:newY + patch_len, :, :] = ver_refo[newY:newY + patch_len, :, :] + frac_vec
# print progress status
self.sta.progress(((x+1)/new_n+.5)*100, self.cfg.params[self.cfg.opt_prnt])
# write image file to hard drive
fp = self.cfg.params[self.cfg.cfn_path].split('.')[0]
img = misc.uint16_norm(ver_refo)
misc.save_img_file(img, os.path.join(fp, 'scheimpflug_' + str(patch_len) + 'px.tiff'))
return True
def main(self):
if self._lfp_path.endswith(('.lfp', '.lfr', '.raw') + tuple('.c.'+str(num) for num in (0, 1, 2, 3))):
# filename and file path from previously decoded data
dp = os.path.splitext(self._lfp_path)[0]
fn = os.path.basename(dp)+'.tiff'
fp = os.path.join(dp, fn)
# load previously generated tiff if present
if os.path.exists(fp):
try:
self._lfp_img = misc.load_img_file(fp)
except TypeError as e:
self.sta.status_msg(e, self.cfg.params[self.cfg.opt_prnt])
self.sta.progress(100, self.cfg.params[self.cfg.opt_prnt])
raise LfpTypeError(e)
except FileNotFoundError as e:
# print status
self.sta.status_msg('File {0} not found'.format(self._lfp_path), self.cfg.params[self.cfg.opt_prnt])
self.sta.progress(100, self.cfg.params[self.cfg.opt_prnt])
raise PlenopticamError(e)
else:
try:
# Lytro type decoding
with open(self._lfp_path, mode='rb') as file:
# print status
self.sta.status_msg('Decode Lytro image file', self.cfg.params[self.cfg.opt_prnt])
self.sta.progress(None, self.cfg.params[self.cfg.opt_prnt])
# LFC and raw type decoding
obj = LfpDecoder(file, self.cfg, self.sta)
if self._lfp_path.endswith(('.lfp', '.lfr') + tuple('.c.'+str(num) for num in (0, 1, 2, 3))):
# LFC type decoding
obj.decode_lfc()
self.cfg.save_json(os.path.join(dp, os.path.basename(dp)+'.json'), json_dict=obj.json_dict)
elif self._lfp_path.endswith('.raw'):
# raw type decoding
obj.decode_raw()
self._lfp_img = obj.rgb_img
del obj
# save bayer image as file
misc.save_img_file(misc.uint16_norm(self._lfp_img), fp, type='tiff')
# print status
self.sta.progress(100, self.cfg.params[self.cfg.opt_prnt])
except FileNotFoundError as e:
# print status
self.sta.status_msg('File {0} not found'.format(self._lfp_path), self.cfg.params[self.cfg.opt_prnt])
self.sta.progress(100, self.cfg.params[self.cfg.opt_prnt])
raise PlenopticamError(e)
else:
try:
# read and decode generic image file type
self._lfp_img = misc.load_img_file(self._lfp_path)
except TypeError as e:
raise LfpTypeError(e)
try:
# try to load json file (if present)
json_dict = self.cfg.load_json(self._lfp_path)
self.cfg.lfpimg = LfpDecoder.filter_json(json_dict)
except:
pass
# write json file
self.cfg.save_params()
return True
def main(self):
if self._lfp_path.lower().endswith(SUPP_FILE_EXT):
# filename and file path from previously decoded data
dp = os.path.splitext(self._lfp_path)[0]
fn = os.path.basename(dp) + '.tiff'
fp = os.path.join(dp, fn)
# load previously generated tiff if present
if os.path.exists(fp):
try:
self._lfp_img = misc.load_img_file(fp)
except FileNotFoundError:
# print status
self.sta.status_msg(
'{0} not found'.format(os.path.basename(
self._lfp_path)),
self.cfg.params[self.cfg.opt_prnt])
self.sta.progress(100, self.cfg.params[self.cfg.opt_prnt])
self.sta.error = True
except TypeError as e:
self.sta.status_msg(e, self.cfg.params[self.cfg.opt_prnt])
self.sta.progress(100, self.cfg.params[self.cfg.opt_prnt])
raise LfpTypeError(e)
else:
try:
# Lytro type decoding
with open(self._lfp_path, mode='rb') as file:
# LFC and raw type decoding
obj = LfpDecoder(file, self.cfg, self.sta)
if self._lfp_path.lower().endswith(SUPP_FILE_EXT[1:]):
# LFC type decoding
obj.decode_lfc()
self.cfg.save_json(os.path.join(
dp,
os.path.basename(dp) + '.json'),
json_dict=obj.json_dict)
elif self._lfp_path.lower().endswith(SUPP_FILE_EXT[0]):
# raw type decoding
obj.decode_raw()
self._lfp_img = obj.rgb_img
del obj
# save bayer image as file
self.sta.status_msg(
msg='Save raw image',
opt=self.cfg.params[self.cfg.opt_prnt])
self.sta.progress(None,
self.cfg.params[self.cfg.opt_prnt])
misc.save_img_file(misc.Normalizer(
self._lfp_img).uint16_norm(),
fp,
file_type='tiff')
self.sta.progress(100,
self.cfg.params[self.cfg.opt_prnt])
except FileNotFoundError:
# print status
self.sta.status_msg(
'{0} not found'.format(os.path.basename(
self._lfp_path)),
self.cfg.params[self.cfg.opt_prnt])
self.sta.progress(100, self.cfg.params[self.cfg.opt_prnt])
self.sta.error = True
except Exception as e:
# unrecognized LFP file type
if not obj.json_dict:
raise LfpTypeError(e)
else:
raise PlenopticamError(e)
else:
try:
# read and decode generic image file type
self._lfp_img = misc.load_img_file(self._lfp_path)
except TypeError as e:
raise LfpTypeError(e)
try:
# try to load json file (if present)
json_dict = self.cfg.load_json(self._lfp_path)
self.cfg.lfpimg = LfpDecoder.filter_json(json_dict)
except:
pass
# write json file
self.cfg.save_params()
return True
def ver_hex_bulge(self, img):
# variable init
img = np.divide(img.astype('float'), img.max(), out=np.zeros_like(img), where=img != 0)
m, n, ch = img.shape if len(img.shape) == 3 else img.shape + (1,)
# j for shifted and i represents truth (unshifted)
i, j = [1, 0] if self.hex_odd else [0, 1]
# gradient between shifted and unshifted (responsive to hex artifacts)
if img[i::2, ...].shape[0] == img[j::2, ...].shape[0]:
dif = img[j::2, ...]-img[i::2, ...]
elif img[j::2, ...].shape[0] == img[i::2, ...].shape[0]+1:
dif = img[j::2, ...][:-1]-img[i::2, ...]
elif img[j::2, ...].shape[0]+1 == img[i::2, ...].shape[0]:
dif = img[j::2, ...] - img[i::2, ...][:-1]
else:
self.sta.error('Image dimension mismatch')
return False
# gradient between unshifted (responsive to real object edges)
edg = img[i::2, ...][:-1]-img[i+2::2, ...]
edg *= 0.75
# absolute values
dif = abs(dif)
edg = abs(edg)
# (don't throw away channel information)
dif = dif.sum(2)
edg = edg.sum(2)
# deduct actual edges
karr = dif - edg if dif.shape[0] == edg.shape[0] else dif[:-1, ...] - edg
karr = np.divide(karr, karr.max(), out=np.zeros_like(karr), where=karr != 0)
karr[karr < 0] = 0
if self.cfg.params[self.cfg.opt_dbug]:
misc.save_img_file(karr, os.path.join(self.cfg.exp_path, 'karr.png'))
# replicate mask to 3 color channels
arr = np.zeros(karr.shape+(3,))
for p in range(ch):
arr[..., p] = karr
# threshold values with insufficient difference
mask = np.divide(arr, arr.max(), out=np.zeros_like(arr), where=arr != 0)
th = 0.1
mask[mask < th] = 0
mask[mask >= th] = 1
# ignore small regions
mask = self.retain_connected(mask, n=4)
# generate mask
if self.cfg.params[self.cfg.opt_dbug]:
full_mask = np.zeros(img.shape)
if full_mask[j::2, ...].shape[0] == mask.shape[0]:
full_mask[j::2, ...] = mask
else:
full_mask[j::2, ...][:-1] = mask
misc.save_img_file(full_mask, os.path.join(self.cfg.exp_path, 'hex_filter_mask.png'), tag=True)
# generate image indicating which pixels were treated
if self.cfg.params[self.cfg.opt_dbug]:
tes = img.copy()
for p in range(ch):
if tes[j::2, ...].shape[0] == mask.shape[0]:
tes[j::2, :, p][mask[..., p] != 0] = tes.max()
else:
tes[j::2, :, p][:-1][mask[..., p] != 0] = tes.max()
tes_out = np.divide(tes, tes.max(), out=np.zeros_like(tes), where=tes != 0)
misc.save_img_file(tes_out, os.path.join(self.cfg.exp_path, 'ident.png'), tag=True)
idxs = np.array(mask.nonzero())
r = 1
valid_idx = np.where((idxs[0] > r) & (idxs[1] > r) & (img[j::2].shape[0]-idxs[0] > r) & (img.shape[1]-idxs[1] > r))[0]
idxs_vals = list(zip(idxs[0][valid_idx], idxs[1][valid_idx], idxs[2][valid_idx]))
res = img.copy()
for idx in idxs_vals:
new_val = np.mean(img[i::2, ...][idx[0]:idx[0]+2, idx[1], idx[2]], axis=0)
res[j::2, ...][idx[0], idx[1], idx[2]] = new_val
if self.cfg.params[self.cfg.opt_dbug]:
misc.save_img_file(res, os.path.join(self.cfg.exp_path, 'hex-corr.png'))
return res