forked from ali1234/vhs-teletext
/
vbi.py
executable file
·354 lines (277 loc) · 11.5 KB
/
vbi.py
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
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
#!/usr/bin/env python
# * Copyright 2011 Alistair Buxton <a.j.buxton@gmail.com>
# *
# * License: This program is free software; you can redistribute it and/or
# * modify it under the terms of the GNU General Public License as published
# * by the Free Software Foundation; either version 3 of the License, or (at
# * your option) any later version. This program is distributed in the hope
# * that it will be useful, but WITHOUT ANY WARRANTY; without even the implied
# * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# * GNU General Public License for more details.
# This is the main data analyser.
import sys
import os
import numpy as np
from scipy.ndimage import gaussian_filter1d as gauss
from scipy.optimize import fminbound
import pylab
from util import paritybytes, setbyte, normalise, hammbytes, allbytes, mrag, notzero
import util
from guess import Guess
import time
import finders
import math
np.seterr(invalid= 'raise')
class Vbi(object):
'''This class represents a line of raw vbi data and all our attempts to
decode it.'''
possible_bytes = [hammbytes]*2 + [paritybytes]*40
def __init__(self, vbi, bitwidth=5.112, gauss_sd=1.1, gauss_sd_offset=2.0,
offset_low = 75.0, offset_high = 119.0,
thresh_low = 1.1, thresh_high = 2.36,
allow_unmatched = True, find=finders.all_headers):
# data arrays
# vbi is the raw line as an array of 2048 floats
self.vbi = vbi
# blurring amounts
self.gauss_sd = gauss_sd
self.gauss_sd_offset = gauss_sd_offset
# Offset range to check for signal drift, in samples.
# The algorithm will check with sub sample accuracy.
# The offset finder is very fast (it uses bisection)
# so this range can be relatively large. But not too
# large or you get false positives.
self.offset_low = offset_low
self.offset_high = offset_high
# black level of the signal
self.black = np.mean(self.vbi[:80])
# Threshold multipliers. The black level of the signal
# is derived from the mean of the area before the VBI
# begins. This is multiplied by the following factors
# to give the low/high thresholds. Anything outside
# this range is assumed to be a 0 or a 1. Tweaking these
# can improve results, but often at a speed cost.
self.thresh_low = self.black*thresh_low
self.thresh_high = self.black*thresh_high
# Allow vbi.py to emitt packet 0's that don't match
# any finder? Set to false when you have finders
# for all headers in the data.
self.allow_unmatched = allow_unmatched
self.finders = find
# vbi packet bytewise
self._mask0 = np.zeros(42, dtype=np.uint8)
self._mask1 = np.zeros(42, dtype=np.uint8)
self.g = Guess(bitwidth=bitwidth)
def find_offset_and_scale(self):
'''Tries to find the offset of the vbi data in the raw samples.'''
# Split into chunks and ensure there is something "interesting" in each
target = gauss(self.vbi, self.gauss_sd_offset)
d = [np.std(target[x:x+128]) < 5.0 for x in range(64, 1440, 128)]
if any(d):
return False
low = 64
high = 256
target = gauss(self.vbi[low:high], self.gauss_sd_offset)
def _inner(offset):
self.g.set_offset(offset)
self.g.update_cri(low, high)
guess_scaled = self.g.convolved[low:high]
mask_scaled = self.g.mask[low:high]
a = guess_scaled*mask_scaled
b = np.clip(target*mask_scaled, self.black, 256)
scale = a.std()/b.std()
b -= self.black
b *= scale
a = np.clip(a, 0, 256*scale)
return np.sum(np.square(b-a))
offset = fminbound(_inner, self.offset_low, self.offset_high)
# call it also to set self.offset and self.scale
return (_inner(offset) < 10)
def make_guess_mask(self):
a = []
for i in range(42*8):
(low, high) = self.g.get_bit_pos(i)
a.append(self.vbi[low:high])
mins = np.array([min(x) for x in a])
maxs = np.array([max(x) for x in a])
avgs = np.array([np.array(x).mean() for x in a])
for i in range(42):
mini = mins[i*8:(i+1)*8]
maxi = maxs[i*8:(i+1)*8]
avgi = avgs[i*8:(i+1)*8]
self._mask0[i] = 0xff
for j in range(8):
if mini[j] < self.thresh_low:
self._mask0[i] &= ~(1<<j)
if maxi[j] > self.thresh_high:
self._mask1[i] |= (1<<j)
tmp = self._mask1 & self._mask0
self._mask0 |= self._mask1
self._mask1 = tmp
def make_possible_bytes(self, possible_bytes):
def masked(b, n):
m0 = util.m0s[self._mask0[n]]
m1 = util.m1s[self._mask1[n]]
m = m0 & m1 & b
if m:
return m
else:
mm0 = m0 & b
mm1 = m1 & b
if len(mm0) < len(mm1):
return mm0 or mm1 or b
else:
return mm1 or mm0 or b
self.possible_bytes = [masked(b,n) for n,b in enumerate(possible_bytes)]
def _deconvolve_make_diff(self, (low, high)):
a = normalise(self.g.convolved)
diff_sq = np.square(a - self.target)
return np.sum(diff_sq)
# an interesting trick I discovered.
# bias the result towards the curent area of interest
return np.sum(diff_sq[:low]) + 2.6*np.sum(diff_sq[low:high]) + np.sum(diff_sq[high:])
def _deconvolve_pass(self, first=0, last=42):
for n in range(first, last):
nb = self.possible_bytes[n]
changed = self.g.set_update_range(n+4, 1)
if len(nb) == 100000:
self.g.set_byte(n, nb[0])
else:
ans = []
for b1 in nb:
self.g.set_byte(n, b1)
ans.append((self._deconvolve_make_diff(changed),b1))
best = min(ans)
self.g.set_byte(n, best[1])
self.g.update_all()
def _deconvolve(self):
for it in range(10):
self._deconvolve_pass()
# if this iteration didn't produce a change in the answer
# then the next one won't either, so stop.
if (self.g.bytes == self._oldbytes).all():
#print it
break
self._oldbytes[:] = self.g.bytes
def _nzdeconvolve(self):
for it in range(10):
ans=[]
changed = self.g.set_update_range(4, 2)
for nb in notzero:
self.g.set_two_bytes(0, nb[0], nb[1])
ans.append((self._deconvolve_make_diff(changed),nb))
best = min(ans)
self.g.set_two_bytes(0, best[1][0], best[1][1])
self._deconvolve_pass(first=2)
# if this iteration didn't produce a change in the answer
# then the next one won't either, so stop.
if (self.g.bytes == self._oldbytes).all():
#print it
break
self._oldbytes[:] = self.g.bytes
def deconvolve(self):
target = gauss(self.vbi, self.gauss_sd)
self.target = normalise(target)
self.make_guess_mask()
self.make_possible_bytes(Vbi.possible_bytes)
self._oldbytes = np.zeros(42, dtype=np.uint8)
self._deconvolve()
packet = "".join([chr(x) for x in self.g.bytes])
F = finders.test(self.finders, packet)
if F:
sys.stderr.write("matched by finder "+F.name+"\n");
sys.stderr.flush()
self.make_possible_bytes(F.possible_bytes)
self._deconvolve()
F.find(self.g.bytes)
packet = F.fixup()
return packet
# if the packet did not match any of the finders then it isn't
# a packet 0 (or 30). if the packet still claims to be a packet 0 it
# will mess up the page splitter. so redo the deconvolution but with
# packet 0 (and 30) header removed from possible bytes.
# note: this doesn't work. i am not sure why. a packet in 63322
# does not match the finders but still passes through this next check
# with r=0. which should be impossible.
((m,r),e) = mrag(self.g.bytes[:2])
if r == 0:
sys.stderr.write("packet falsely claimed to be packet %d\n" % r);
sys.stderr.flush()
if not self.allow_unmatched:
self._nzdeconvolve()
packet = "".join([chr(x) for x in self.g.bytes])
# if it's a link packet, it is completely hammed
elif r == 27:
self.make_possible_bytes([hammbytes]*42)
self._deconvolve()
packet = "".join([chr(x) for x in self.g.bytes])
return packet
def process_file((inname, outname)):
print inname, outname
try:
data = np.fromstring(file(inname).read(), dtype=np.uint8)
outfile = file(outname, 'wb')
for line in range(32):
offset = line*2048
vbiraw = data[offset:offset+2048]
v = Vbi(vbiraw)
tmp = v.find_offset_and_scale()
if tmp:
outfile.write(v.deconvolve())
else:
outfile.write("\xff"*42)
outfile.close()
except IOError:
pass
def test_file(outfile):
return os.path.isfile(outfile) and (os.path.getsize(outfile) == (42 * 32))
def list_files(inputpath, outputpath, first, count, skip):
for frame in range(first, first+count, skip):
frame = "%08d" % frame
if test_file(outputpath + '/' + frame + '.t42'):
#print "skipping %s\n" % (outputpath + '/' + frame + '.t42')
pass
else:
yield (inputpath + '/' + frame + '.vbi',
outputpath + '/' + frame + '.t42')
if __name__ == '__main__':
import multiprocessing
from multiprocessing.pool import IMapIterator, Pool
import cProfile
import pstats
def wrapper(func):
def wrap(self, timeout=None):
# Note: the timeout of 1 googol seconds introduces a rather subtle
# bug for Python scripts intended to run many times the age of the universe.
return func(self, timeout=timeout if timeout is not None else 1e100)
return wrap
IMapIterator.next = wrapper(IMapIterator.next)
try:
path = sys.argv[1]
except:
print "Usage:", sys.argv[0], "<path> [<first> <count>]\n"
print " path: directory with VBI files to process"
print " first: first file to process"
print " count: maximum number of files to process\n"
exit(-1)
try:
first = int(sys.argv[2], 10)
count = int(sys.argv[3], 10)
skip = int(sys.argv[4], 10)
except:
first = 0
count = 10000000
skip = 1
if not os.path.isdir(path+'/t42/'):
os.makedirs(path+'/t42/')
if 1:
p = Pool(multiprocessing.cpu_count())
it = p.imap(process_file, list_files(path+'/vbi/', path+'/t42/', first, count, skip), chunksize=1)
for i in it:
pass
else: # single thread mode for debugging
def doit():
map(process_file, list_files(path+'/vbi/', path+'/t42/', first, count, skip))
cProfile.run('doit()', 'myprofile')
p = pstats.Stats('myprofile')
p.sort_stats('cumulative').print_stats(50)