-
Notifications
You must be signed in to change notification settings - Fork 5
/
init.lua
1257 lines (1179 loc) · 41.5 KB
/
init.lua
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
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
--
-- Note: this bit of code is a simple wrapper around the OpenCV library
-- http://opencv.willowgarage.com/
--
-- For now, it contains wrappers for:
-- + opencv.GetAffineTransform() [lua] --> cvGetAffineTransform [C/C++]
-- + opencv.WarpAffine() [lua] --> cvWarpAffine [C/C++]
-- + opencv.EqualizeHist() [lua] --> cvEqualizeHist [C/C++]
-- + opencv.Canny() [lua] --> cvCanny [C/C++]
-- + opencv.CornerHarris() [lua] --> cvCornerHarris [C/C++]
--
-- + opencv.CalcOpticalFlow() [lua] -->
-- - cvCalcOpticalFlowBM
-- - cvCalcOpticalFlowHS
-- - cvCalcOpticalFlowLK
--
-- + opencv.GoodFeaturesToTrack() [lua] --> cvGoodFeaturesToTrack [C/C++]
--
-- Wrapper: Clement Farabet.
-- Additional functions: GoodFeatures...(),PLK,etc.: Marco Scoffier
-- Adapted for torch7: Marco Scoffier
--
require 'torch'
require 'dok'
require 'image'
opencv = {}
-- load C lib
require 'libopencv'
function opencv.StereoCorrespondenceBM(...)
local _, left, right, minDisparity, numberOfDisparities, textureThreshold = dok.unpack(
{...},
'opencv.StereoCorrespondenceBM',
[[Implements opencv StereoCorrespondenceBM using block matching algorithm
return Tensor of computed disparities]],
{arg='left', type='torch.Tensor',
help='single channel left image', req=true},
{arg='right', type='torch.Tensor',
help='single channel right image', req=true},
{arg='minDisparity',type='number',
help='minimum disparity', default=1},
{arg='numberOfDisparities',type='number',
help='maximum disparity - minimum disparity must be a multiple of 16', default=64},
{arg='textureThreshold',type='number',
help='areas with no texture are ignored', default=10}
)
-- check left
local leftImg = torch.Tensor(left)
if left:nDimension() == 2 then
leftImg = leftImg:resize(1,left:size(1),left:size(2))
elseif leftImg:nDimension() == 3 and leftImg:size(1) == 3
then
leftImg=image.rgb2y(left)
elseif leftImg:size(1) ~= 1 then
xerror([[ *** ERROR: opencv.StereoCorrespondenceBM
left img must be RBG or grey scale]])
end
-- check right
local rightImg = torch.Tensor(right)
if right:nDimension() == 2 then
rightImg = rightImg:resize(1,right:size(1),right:size(2))
elseif rightImg:nDimension() == 3 and rightImg:size(1) == 3
then
rightImg=image.rgb2y(right)
elseif rightImg:size(1) ~= 1 then
xerror([[ *** ERROR: opencv.StereoCorrespondenceBM
right img must be RBG or grey scale]])
end
local dest = torch.Tensor()
left.libopencv.StereoCorrespondenceBM(leftImg,rightImg,dest,
minDisparity, numberOfDisparities,
textureThreshold)
return dest
end
function opencv.StereoCorrespondenceGC(...)
local _, left, right, maxIters, numberOfDisparities = dok.unpack(
{...},
'opencv.StereoCorrespondenceGC',
[[Implements opencv StereoCorrespondenceBM using block matching algorithm
return left disparities, right disparities]],
{arg='left', type='torch.Tensor',
help='single channel left image', req=true},
{arg='right', type='torch.Tensor',
help='single channel right image', req=true},
{arg='maxIters',type='number',
help='maximum # of iterations', default=2},
{arg='numberOfDisparities',type='number',
help='maximum disparity - minimum disparity must be a multiple of 16', default=16}
)
-- check left
local leftImg = torch.Tensor(left)
if left:nDimension() == 2 then
leftImg = leftImg:resize(1,left:size(1),left:size(2))
elseif leftImg:nDimension() == 3 and leftImg:size(1) == 3
then
leftImg=image.rgb2y(left)
elseif leftImg:size(1) ~= 1 then
xerror([[ *** ERROR: opencv.StereoCorrespondenceGC
left img must be RBG or grey scale]])
end
-- check right
local rightImg = torch.Tensor(right)
if right:nDimension() == 2 then
rightImg = rightImg:resize(1,right:size(1),right:size(2))
elseif rightImg:nDimension() == 3 and rightImg:size(1) == 3
then
rightImg=image.rgb2y(right)
elseif rightImg:size(1) ~= 1 then
xerror([[ *** ERROR: opencv.StereoCorrespondenceGC
right img must be RBG or grey scale]])
end
local Ldisp = torch.Tensor()
local Rdisp = torch.Tensor()
left.libopencv.StereoCorrespondenceGC(leftImg,rightImg,Ldisp,Rdisp,
maxIters, numberOfDisparities)
return Ldisp,Rdisp
end
-- Canny
function opencv.Canny(...)
local _, source, percent, low_threshold, high_threshold, blursize, aperturesize = dok.unpack(
{...},
'opencv.Canny',
[[Implements the Canny algorithm for edge detection.
return Tensor of edges, low_threshold,high_threshold (thresholds used for the computation)]],
{arg='source', type='torch.Tensor',
help='image in which to perform edge detection', req=true},
{arg='percent',type='number',
help='determine automatically low and high threshold'},
{arg='low',type='number',
help=[[The smallest value between low and high is used
for edge linking, the largest value is used to find the initial segments
of strong edges. note that in opencv the pixels are in [0 255].
also note that using percent param will ignore low and high values ]], default=50},
{arg='high',type='number',
help='cf. low',default=150},
{arg='blursize',type='number',
help='Guassian blur kernel size (<1 means no blurring)', default=0},
{arg='aperturesize',type='number',
help='Sobel aperture size', default=3}
)
local img = source
if source:size(1) == 3 then
print('WARNING: opencv.Canny converting image to grey')
img=image.rgb2y(source)
elseif source:size(1) ~= 1 then
xerror(' *** ERROR: opencv.Canny works only on RBG or grey img')
end
if blursize>1 and blursize % 2 == 0 then
print('WARNING: blursize (Guassian blur kernel size) must be odd')
blursize = blursize -1
end
if aperturesize % 2 == 0 or aperturesize < 1 or aperturesize > 7 then
print('WARNING: aperturesize (Sobel kernel size) must be odd >= 1 and <= 7')
aperturesize = math.max(1,math.min(aperturesize -1,7))
end
local dest = torch.Tensor():resizeAs(img)
low_threshold,high_threshold = img.libopencv.Canny(img,dest,low_threshold,high_threshold,blursize,aperturesize,percent)
return dest,low_threshold,high_threshold
end
function opencv.GetAffineTransform(...)
local args, points_src, points_dst = dok.unpack(
{...},
'opencv.GetAffineTransform',
[[Calculates the affine transform from 3 corresponding points. ]],
{arg='points_src',type='torch.Tensor',
help='source points', req=true},
{arg='points_dst',type='torch.Tensor',
help='destination points', req=true}
)
local warp = torch.Tensor()
warp.libopencv.GetAffineTransform(points_src,points_dst,warp)
return warp
end
-- test function:
function opencv.GetAffineTransform_testme()
src = torch.Tensor(3,2)
dst = torch.Tensor(3,2)
src[1][1]=0
src[1][2]=0
src[2][1]=511
src[2][2]=0
src[3][1]=0
src[3][2]=511
dst[1][1]=0
dst[1][2]=512*0.25
dst[2][1]=512*0.9
dst[2][2]=512*0.15
dst[3][1]=512*0.1
dst[3][2]=512*0.75
warp = opencv.GetAffineTransform(src,dst)
print('Warp matrix:')
print(warp)
end
-- WarpAffine
function opencv.WarpAffine(...)
local _, source,warp = dok.unpack(
{...},
'opencv.WarpAffine',
[[Implements the affine transform which allows the user to warp,
stretch, rotate and resize an image.]],
{arg='source', type='torch.Tensor',
help='image in which to perform Histogram Equalization', req=true},
{arg='warp', type='torch.Tensor',
help='2x3 transformation matrix', req=true}
)
local img = source
if warp:size(1) ~= 2 or warp:size(2) ~= 3 then
xerror(' *** ERROR: opencv.WarpAffine warp Tensor must be 2x3')
end
local dest = torch.Tensor():resizeAs(img)
img.libopencv.WarpAffine(img,dest,warp)
return dest
end
-- test function:
function opencv.WarpAffine_testme(img)
if not img then
img = image.lena()
image.display{image=img,legend='Original image'}
end
local src = torch.Tensor(3,2)
local dst = torch.Tensor(3,2)
src[1][1]=0
src[1][2]=0
src[2][1]=511
src[2][2]=0
src[3][1]=0
src[3][2]=511
dst[1][1]=0
dst[1][2]=512*0.15
dst[2][1]=512*0.9
dst[2][2]=512*0.05
dst[3][1]=512*0.1
dst[3][2]=512*0.75
local warp = opencv.GetAffineTransform(src,dst)
print('warp',warp)
local warpImg = opencv.WarpAffine(img,warp)
image.display{image=warpImg,legend='Warped image'}
end
-- EqualizeHist
function opencv.EqualizeHist(...)
local _, source = dok.unpack(
{...},
'opencv.EqualizeHist',
'Implements the Histogram Equalization algorithm.',
{arg='source', type='torch.Tensor',
help='image in which to perform Histogram Equalization', req=true}
)
local img = source
if source:size(1) == 3 then
print('WARNING: opencv.EqualizeHist converting image to grey')
img=image.rgb2y(source)
elseif source:size(1) ~= 1 then
xerror(' *** ERROR: opencv.EqualizeHist works only on RBG or grey img')
end
local dest = torch.Tensor():resizeAs(img)
img.libopencv.EqualizeHist(img,dest)
return dest
end
-- CornerHarris
function opencv.CornerHarris(...)
local _, img, blocksize, aperturesize, k = dok.unpack(
{...},
'opencv.CornerHarris',
'Computes the Harris Corner features of an image the input image will be converted to a WxHx1 tensor',
{arg='img', type='torch.Tensor',
help='image in which to detect Haar points', req=true},
{arg='blocksize',type='number',
help='neighborhood size', default=9},
{arg='aperturesize',type='number',
help='Sobel aperture size', default=3},
{arg='k',type='number',
help='the Harris detector free parameter',default=0.04}
)
local img = img
if img:size(1) > 1 then
print('WARNING: computing harris corners on first feature')
img=img:narrow(1,1,1)
end
if aperturesize % 2 == 0 then
print('WARNING: aperturesize (Sobel kernel size) must be odd and not larger than 31')
aperturesize = aperturesize -1
end
local harris = torch.Tensor():resizeAs(img)
img.libopencv.CornerHarris(img,harris,blocksize,aperturesize,k)
return harris
end
function opencv.imgL()
return image.load(sys.concat(sys.fpath(), 'img1.jpg'))
end
function opencv.imgR()
return image.load(sys.concat(sys.fpath(), 'img2.jpg'))
end
-- test function:
function opencv.CornerHarris_testme(img)
if not img then
img = opencv.imgL()
image.display{image=img,legend='Original image (Left)'}
end
local harris = opencv.CornerHarris(img,5,3,0.05)
image.display{image=harris,legend='Harris Corners (Left)'}
end
-- OpticalFlow:
function opencv.CalcOpticalFlow(...)
local args, pair, method, block_w, block_h,
shift_x, shift_y, window_w, window_h,
lagrangian, iterations, autoscale,
raw, reuse, flow_x, flow_y = dok.unpack(
{...},
'opencv.CalcOpticalFlow',
[[
Computes the optical flow of a pair of images, and returns 4 maps:
the flow field intensities, the flow field directions, and
the raw X and Y components
The flow field is computed using one of 3 methods:
Block Matching (BM), Lucas-Kanade (LK) or Horn-Schunck (HS).
The input images must be a pair of WxHx1 tensors.
]],
{arg='pair', type='table',
help='a pair of images (2 WxHx1 tensor)', req=true},
{arg='method', type='string',
help='method used: BM | HS | LK', default='BM'},
{arg='block_w', type='number',
help='matching block width (BM+LK)', default=9},
{arg='block_h', type='number',
help='matching block height (BM+LK)', default=9},
{arg='shift_x', type='number',
help='shift step in x (BM only)', default=4},
{arg='shift_y', type='number',
help='shift step in y (BM only)', default=4},
{arg='window_w', type='number',
help='matching window width (BM only)', default=30},
{arg='window_h', type='number',
help='matching window height (BM only)', default=30},
{arg='lagrangian', type='number',
help='lagrangian multiplier (HS only)', default=1},
{arg='iterations', type='number',
help='nb of iterations (HS only)', default=5},
{arg='autoscale', type='boolean',
help='auto resize results', default=true},
{arg='raw', type='boolean',
help='if set, returns the raw X,Y fields', default=false},
{arg='reuse', type='boolean',
help='reuse last flow computed (HS+BM)', default=false},
{arg='flow_x', type='torch.Tensor',
help='existing (previous) X-field (WxHx1 tensor)'},
{arg='flow_y', type='torch.Tensor',
help='existing (previous) Y-field (WxHx1 tensor)'}
)
if pair[1]:nDimension() ~= 3 then
dok.error('inconsistent input size'..args.usage,
'opencv.CalcOpticalFlow')
end
local imageP = pair[1]
local imageN = pair[2]
if imageP:size(1) > 1 then
print('WARNING: computing flow on first feature')
imageP=imageP:narrow(1,1,1)
end
if imageN:size(1) > 1 then
print('WARNING: computing flow on first feature')
imageN=imageN:narrow(1,1,1)
end
local flow_x = flow_x or torch.Tensor()
local flow_y = flow_y or torch.Tensor()
if method == 'BM' then
imageP.libopencv.CalcOpticalFlow(imageN, imageP, flow_x, flow_y, 1,
block_w, block_h,
shift_x, shift_y,
window_w, window_h,
reuse)
elseif method == 'LK' then
imageP.libopencv.CalcOpticalFlow(imageN, imageP, flow_x, flow_y, 2,
block_w, block_h)
elseif method == 'HS' then
imageP.libopencv.CalcOpticalFlow(imageN, imageP, flow_x, flow_y, 3,
lagrangian, iterations,
-1,-1,-1,-1,
reuse)
else
print('Unkown method')
error(args.usage)
end
if raw then
if autoscale then
local flow_x_s = torch.Tensor():resizeAs(imageP)
local flow_y_s = torch.Tensor():resizeAs(imageP)
print('flow_x_s:size()',flow_x_s:size())
print('flow_y_s:size()',flow_y_s:size())
image.scale(flow_x, flow_x_s, 'simple')
image.scale(flow_y, flow_y_s, 'simple')
return flow_x_s, flow_y_s
else
return flow_x, flow_y
end
else
local flow_norm = torch.Tensor()
local flow_angle = torch.Tensor()
-- compute norm:
local x_squared = torch.Tensor():resizeAs(flow_x):copy(flow_x):cmul(flow_x)
flow_norm:resizeAs(flow_y):copy(flow_y):cmul(flow_y):add(x_squared):sqrt()
-- compute angle:
flow_angle:resizeAs(flow_y):copy(flow_y):cdiv(flow_x):abs():atan():mul(180/math.pi)
flow_angle:map2(flow_x, flow_y,
function(h,x,y)
if x == 0 and y >= 0 then
return 90
elseif x == 0 and y <= 0 then
return 270
elseif x >= 0 and y >= 0 then
-- all good
elseif x >= 0 and y < 0 then
return 360 - h
elseif x < 0 and y >= 0 then
return 180 - h
elseif x < 0 and y < 0 then
return 180 + h
end
end)
if autoscale then
local flow_norm_s = torch.Tensor():resizeAs(imageP)
local flow_angle_s = torch.Tensor():resizeAs(imageP)
local flow_x_s = torch.Tensor():resizeAs(imageP)
local flow_y_s = torch.Tensor():resizeAs(imageP)
image.scale(flow_angle, flow_angle_s, 'simple')
image.scale(flow_norm, flow_norm_s, 'simple')
image.scale(flow_x, flow_x_s, 'simple')
image.scale(flow_y, flow_y_s, 'simple')
return flow_norm_s, flow_angle_s, flow_x_s, flow_y_s
else
return flow_norm, flow_angle, flow_x, flow_y
end
end
end
function opencv.display(...)
local args = {}
xlua.unpack_class(args, {...}, 'opencv.display',
'displays a single image, with optional parameters',
{arg='image', type='torch.Tensor | table',
help='image or table of images 3xHxW', req=true},
{arg='min', type='number', help='lower-bound for range', default=nil},
{arg='max', type='number', help='upper-bound for range', default=nil},
{arg='win', type='string', help='window legend (and descriptor)', default=nil}
)
function cloneImg(src, dst)
if src:size(1) == 1 then
src = src[1]
end
if src:nDimension() == 2 then
dst[1]:copy(src)
dst[2]:copy(src)
dst[3]:copy(src)
else
dst:copy(src)
end
end
local image
if type(args.image) == 'table' then
local n = #args.image
local h = args.image[1]:size(2)
local w = args.image[1]:size(3)
image = torch.Tensor(3, h, n*w)
for i = 1,n do
cloneImg(args.image[i], image:narrow(3, (i-1)*w+1, w))
end
else
local h, w
if args.image:nDimension() == 2 then
h = args.image:size(1)
w = args.image:size(2)
else
h = args.image:size(2)
w = args.image:size(3)
end
image = torch.Tensor(3, h, w)
cloneImg(args.image, image)
end
if not args.min then args.min = image:min() end
if not args.max then args.max = image:max() end
image:add(-args.min)
image:div(args.max-args.min)
opencv.windownames = opencv.windownames or {}
if not args.win then
local i = 1
while opencv.windownames[string.format("window %d", i)] do
i = i + 1
end
args.win = string.format("window %d", i)
opencv.windownames[args.win] = true
end
image.libopencv.display(image, args.win)
return args.win
end
-- testers:
function opencv.CalcOpticalFlow_testme(img1, img2)
local img1 = img1
local img2 = img2
if not img1 then
img1 = opencv.imgL()
image.display{image=img1,legend='Original image (Left)'}
end
if not img2 then
img2 = opencv.imgR()
image.display{image=img2,legend='Original image (Right)'}
end
img1 = image.scale(img1,img1:size(3)/2,img1:size(2)/2)
img2 = image.scale(img2,img2:size(3)/2,img2:size(2)/2)
local methods = {'LK', 'HS', 'BM'}
for i,method in ipairs(methods) do
print(i,method)
local norm, angle, flow_x, flow_y =
opencv.CalcOpticalFlow{pair={img1,img2}, method=method}
local hsl = torch.Tensor(3,img1:size(2), img1:size(3))
hsl:select(1,1):copy(angle):div(360)
hsl:select(1,2):copy(norm):div(math.max(norm:max(),1e-2))
hsl:select(1,3):fill(0.5)
local rgb = image.hsl2rgb(hsl)
image.display{image={img1,img2,rgb},
legend='cvOpticalFLow, method = ' .. method}
image.display{image={norm,angle,flow_x,flow_y},
scaleeach=true,
legend='cvOpticalFLow, method = ' .. method}
end
end
-- GoodFeaturesToTrack
opencv.GoodFeaturesToTrack
= function(...)
local args, image, count, quality, min_distance, win_size, mask =
dok.unpack(
{...},
'opencv.GoodFeaturesToTrack',
[[
Computes the GoodFeatures algorithm of opencv.
+ returns a points tensor (npoints x 2) of the pixel positions
(x,y in image space) of the features.
]],
{arg='image', type='torch.Tensor',
help='image in which to detect Good Feature points',req=true},
{arg='count',type='number',
help='number of points to return', default=500},
{arg='quality',type='number',
help='quality', default=0.01},
{arg='min_distance',type='number',
help='min spatial distance (in pixels) between returned feature points', default=10},
{arg='win_size',type='number',
help='window size over which to run heuristics', default=10},
{arg='mask', type='torch.Tensor',
help="a tensor of 0 and 1s which specifies a region of the image on which to run the goodfeatures",default=nil}
)
local img = image
local points = torch.Tensor(2,count)
img.libopencv.GoodFeaturesToTrack(img,
points,
count,
quality,
min_distance,
win_size,
mask)
return points
end
-- testers:
function opencv.GoodFeaturesToTrack_testme(img)
if not img then
img = opencv.imgL()
end
sys.tic()
local pts = opencv.GoodFeaturesToTrack{image=img,count=125}
local s = sys.toc()
print("Found "..pts:size(1).." points in "..s.." secs")
end
opencv.CalcOpticalFlowPyrLK
= function(...)
local args, image_from, image_to =
dok.unpack(
{...},
'opencv.CalcOpticalFlowPyrLK',
[[
Computes the Pyramidal Lucas-Kanade optical flow algorithm of opencv.
+ input two images
+ returns a points tensor of the pixel positions of the features
and a copy of the input image with red lines indicating the flow from
the interest points
]],
{arg='image_from', type='torch.Tensor',
help='image in which calculate from flow',req=true},
{arg='image_to', type='torch.Tensor',
help='image in which calculate to flow',req=true}
)
-- need to clean this up can be internal to C function
local flowx = torch.Tensor(image_from:size(1),image_from:size(2)):zero()
local flowy = torch.Tensor(image_from:size(1),image_from:size(2)):zero()
local points = torch.Tensor(500,2)
local image_out = torch.Tensor():resizeAs(image_to):copy(image_to)
image_from.libopencv.CalcOpticalFlowPyrLK(image_from,image_to,flowx,flowy,points,image_out)
return points, image_out
end
function opencv.CalcOpticalFlowPyrLK_testme(imgL,imgR)
if not imgL then
imgL = opencv.imgL()
end
if not imgR then
imgR = opencv.imgR()
end
local points, image_out = opencv.CalcOpticalFlowPyrLK(imgL,imgR)
image.display(image_out)
end
function opencv.LowLevelConversions_testme(img)
if not img then
img = opencv.imgL()
end
local imgn = img:narrow(1,1,1)
local dst = torch.Tensor()
print('Testing torch>IPL8U ... 1 Channel')
img.libopencv.test_torch2IPL8U(imgn,dst)
local err = (imgn-dst):max()
if err > 1/255 then
print (' ERROR '..err)
else
print (' OK')
end
print('Testing torch>IPL8U ... 3 Channels')
img.libopencv.test_torch2IPL8U(img,dst)
local err = (img-dst):max()
if err > 1/255 then
print (' ERROR '..err)
else
print (' OK')
end
print('Testing torch>IPL32F ... 1 Channel')
dst = torch.Tensor()
img.libopencv.test_torch2IPL32F(imgn,dst)
err = (imgn-dst):max()
if err > 0 then
print (' ERROR '..err)
else
print (' OK')
end
print('Testing torch>IPL32F ... 3 Channels')
dst = torch.Tensor()
img.libopencv.test_torch2IPL32F(img,dst)
err = (img-dst):max()
if err > 0 then
print (' ERROR '..err)
else
print (' OK')
end
end
-- Pyramidal Lucas-Kanade
opencv.TrackPyrLK
= function(...)
local args, pair, points_in, points_out, win_size = dok.unpack(
{...},
'opencv.TrackPyrLK',
[[
Runs pyramidal Lucas-Kanade, on two input images and a set of
points which are meant to be tracked.
Returns 3 tensors:
- points : the npoints x 2 list of matching points in the second image
- feature_found : binary npoints whether the feature was found
- feature_error : how close the found feature matches original
]],
{arg='pair', type='table',
help='a pair of images (2 WxHx1 tensor)', req=true},
{arg='points_in',type='torch.Tensor',
help='points to track', req=true},
{arg='points_out',type='torch.Tensor',
help='tensor to return location of tracked points in output'},
{arg='win_size',type='number',
help='over how large of a window can the LK track', default= 25}
)
if not points_out then
points_out = torch.Tensor():resizeAs(points_in):zero()
end
local feature_found = torch.Tensor(points_in:size(1)):zero()
local feature_error = torch.Tensor(points_in:size(1)):zero()
pair[1].libopencv.TrackPyrLK(pair[1], pair[2],
points_in, points_out, win_size,
feature_found, feature_error)
return points_out, feature_found, feature_error
end
opencv.drawFlowlinesOnImage
= function (...)
local args, pair, image, color, mask = dok.unpack(
{...},
'opencv.drawFlowlinesOnImage',
[[ utility to visualize sparse flows ]],
{arg='pair', type='table',
help='a pair of point tensors (2 nPointsx(2 or 3) tensor)', req=true},
{arg='image', type='torch.Tensor',
help='image on which to draw the flowlines', req=true},
{arg='color', type='torch.Tensor',
help='color of flow line eg. R = [255,0,0]'},
{arg='mask', type='torch.Tensor',
help='mask tensor 1D npoints 0 when not to draw point'}
)
-- default color is red
if not color then
color = torch.Tensor(3):zero()
color[1] = 255
end
pair[1].libopencv.drawFlowlinesOnImage(pair[1],pair[2],image,color,mask)
end
opencv.circlePoints
= function (...)
local args, points, image, color, size = dok.unpack(
{...},
'opencv.circlePoints',
[[ draw circles around interest points ]],
{arg='points', type='torch.Tensor',
help='a point tensor (nPointsx(2or3) tensor)', req=true},
{arg='image', type='torch.Tensor',
help='image on which to draw the circles', req=true},
{arg='color', type='torch.Tensor',
help='color of flow line eg. R = [255,0,0]'},
{arg='size', type='number',
help='size of the circles to draw'}
)
-- default color is red
if not color then
color = torch.Tensor(3):zero()
color[1] = 255
end
if not size then
size = 10
end
points.libopencv.circlePoints(points,image,color,size)
end
opencv.drawPoly
= function (...)
local args, points, image, fill, color = dok.unpack(
{...},
'opencv.drawPoly',
[[ draw polygon on an image ]],
{arg='points', type='torch.Tensor',
help='a point tensor (nPointsx(2) tensor)', req=true},
{arg='image', type='torch.Tensor',
help='image on which to draw the polygon', req=true},
{arg="fill", type="boolean",
help="fill polygon or draw outline"},
{arg='color', type='torch.Tensor',
help='color of fill or outline eg. RGB = [255,0,0] or RGBA = [255,255,255,50]'}
)
-- default color is red
if not color then
color = torch.Tensor(3):zero()
color[1] = 255
end
local fillint
if not fill then
fillint = 0
else
fillint = 1
end
points.libopencv.drawPoly(points,image,fillint,color)
end
function opencv.TrackPyrLK_testme(imgL,imgR)
if not imgL then
imgL = opencv.imgL()
end
if not imgR then
imgR = opencv.imgR()
end
local ptsin = opencv.GoodFeaturesToTrack{image=imgL,count=imgL:nElement()}
local ptsout = opencv.TrackPyrLK{pair={imgL,imgR},points_in=ptsin}
opencv.drawFlowlinesOnImage({ptsin,ptsout},imgR)
image.display{image={imgL,imgR},
legend='opencv: Optical Flow Pyramidal LK Tracking',
win_w=imgL:size(1)*2,win_h=imgL:size(2)}
end
function opencv.TrackPyrLK_wmask_testme(imgL,imgR,mask)
if not imgL then
imgL = opencv.imgL()
end
if not imgR then
imgR = opencv.imgR()
end
if not mask then
mask = torch.Tensor(imgL:size(2),imgL:size(3)):fill(0)
mask:narrow(1,200,200):narrow(2,100,250):fill(1)
end
local ptsin = opencv.GoodFeaturesToTrack{image=imgL,
count=imgL:nElement(),
mask=mask}
local ptsout = opencv.TrackPyrLK{pair={imgL,imgR},points_in=ptsin}
opencv.drawFlowlinesOnImage({ptsin,ptsout},imgR)
image.display{image={imgL,imgR},
legend='opencv: Optical Flow Pyramidal LK Tracking',
win_w=imgL:size(1)*2,win_h=imgL:size(2)}
end
-- opencv.smoothVoronoi
-- = function (...)
-- local args, points, data, output = dok.unpack(
-- {...},
-- 'opencv.smoothVoronoi',
-- [[ dense interpolation of sparse flows ]],
-- {arg='points', type='torch.Tensor',
-- help='nPoints x 2 tensor -- locations', req=true},
-- {arg='data', type='torch.Tensor',
-- help='nPoints x n tensor -- data', req=true},
-- {arg='output', type='torch.Tensor',
-- help='bounding rectangle in which dense flows will be stored'}
-- )
-- if not output then
-- local width = points:select(2,1):max()
-- local height = points:select(2,2):max()
-- output = torch.Tensor(data:size(2),height,width)
-- end
-- sys.tic()
-- data.libopencv.smoothVoronoi(points,data,output)
-- print("time to compute dense flow: ",sys.toc())
-- return output
-- end
-- function opencv.smoothVoronoi_testme(imgL,imgR)
-- if not imgL then
-- imgL = opencv.imgL()
-- end
-- if not imgR then
-- imgR = opencv.imgR()
-- end
-- ptsin = opencv.GoodFeaturesToTrack{image=imgL,count=imgL:nElement()}
-- ptsout = opencv.TrackPyrLK{pair={imgL,imgR},points_in=ptsin}
-- output = opencv.smoothVoronoi(ptsout,ptsout-ptsin)
-- image.display{image={output:select(1,1),output:select(1,2)}}
-- end
opencv.findFundamental
= function (...)
local args, points1, points2 = dok.unpack(
{...},
'opencv.FindFundamental',
[[ find fundamental matrix in 2 sets of points]],
{arg='points1', type='torch.Tensor',
help='nPoints x 2 tensor -- locations', req=true},
{arg='points2', type='torch.Tensor',
help='nPoints x 2 tensor -- locations', req=true}
)
if not output then
output = torch.Tensor(3,3)
end
if not status then
status = torch.Tensor(points1:size(1))
end
points1.libopencv.FindFundamental(points1,points2,output,status)
return output,status
end
function opencv.findFundamental_testme(imgL,imgR)
if not imgL then
imgL = opencv.imgL()
end
if not imgR then
imgR = opencv.imgR()
end
sys.tic()
local ptsin = opencv.GoodFeaturesToTrack{image=imgL,count=imgL:nElement()}
local ptsout = opencv.TrackPyrLK{pair={imgL,imgR},points_in=ptsin}
print("time to compute good features: ",sys.toc())
sys.tic()
local matrix,status = opencv.findFundamental{points1=ptsin,points2=ptsout}
print("time to compute fundamental matrix: ",sys.toc())
print(matrix)
end
opencv.findEssential
= function (...)
local args, fundamental, calibration = dok.unpack(
{...},
'opencv.FindEssential',
[[ find essential matrix from fundamental and calibration]],
{arg='fundamental', type='torch.Tensor',
help='3x3 fundamental matrix', req=true},
{arg='calibration', type='torch.Tensor',
help='3x3 tensor -- camera calibration', req=true}
)
return calibration:transpose(1,2) * fundamental * calibration
end
function opencv.findEssential_testme(imgL,imgR)
if not imgL then
imgL = opencv.imgL()
end
if not imgR then
imgR = opencv.imgR()
end
sys.tic()
local ptsin = opencv.GoodFeaturesToTrack{image=imgL,count=imgL:nElement()}
local ptsout = opencv.TrackPyrLK{pair={imgL,imgR},points_in=ptsin}
print("time to compute good features: ",sys.toc())
sys.tic()
local fundmat,status = opencv.findFundamental{points1=ptsin,points2=ptsout}
print("time to compute fundamental matrix: ",sys.toc())
local k = torch.Tensor(3,3):fill(0)
k[1][1] = 602 -- focal length in pixels
k[2][2] = 602 -- focal length in pixels
k[1][3] = 1280/2 -- center width
k[2][3] = 720/2 -- center height
k[3][3] = 1
sys.tic()
local essenmat = opencv.findEssential(fundmat, k)
print("time to compute essential matrix: ",sys.toc())
print(essenmat)
end
opencv.getExtrinsicsFromEssential
= function (...)
local args, essential, point = dok.unpack(
{...},
'opencv.getExtrinsicsFromEssential',
[[ get camera extrinsic matrix from essential matrix and a single point]],
{arg='essential', type='torch.Tensor',
help='3x3 essential matrix', req=true},
{arg='point', type='torch.Tensor',
help='3x1 tensor -- 3D point', req=true}
)
-- u*torch.diag(s)*v:t()
local u,s,v = torch.svd(essential)
-- if ((math.abs(s[1] - s[2]) > 0.1) or math.abs(s[3]) > 0.01) then
-- dok.error("bad essential matrix")
-- end
print("U")
print(u)
print("S - diagonal")
print(s)
print("V -- which you transpose")
print(v)
local w = torch.Tensor(3,3)
w[1][2] = -1
w[2][1] = 1
w[3][3] = 1
local z = torch.Tensor(3,3)
z[1][2] = 1
z[2][1] = -1
extr = torch.Tensor(3,4)
extr:narrow(2,1,3):copy(u*w*v:t())
extr:narrow(2,3,1):copy(u:narrow(1,3,1))
print(extr)
end
function opencv.getExtrinsicsFromEssential_testme(imgL,imgR)
if not imgL then