Captcha Recognizer

Nemesis

Our Projects

Extra

Controls

A captcha from filebeam.com looks like this:

This captcha is as easy as the captcha from netload.in, differences are the background, the font used and the number of digits. The background has random noise but it's static. Making a sample with the background and using it to remove the background when recognizing this captcha would be a sollution but not a good one. Instead of removing this background it's better to remove bagrounds of this kind in case they want to change something about it. We do the histogram for this image.

Highest frequencies are the background noise so they are removed. There is one exception - last digit is always black, same colour used for the bounding rectangle so the colour of the rectangle is avoided.

We adjust the histogram to remove the noise from background and from the digits that need to be identified.

We remove isolated pixels and pixel groups.

Doing a pattern search would require too many patterns and the error rate would still be high. Instead, a line analysis is being done.

For a more accurate identification, the numbers are re-scaled to fit in a 16x16 matrix (see matrix.txt).

0001000000000000 0111111000000000 1110011100000000 1000011110000000 0000000110000000 0000000110000000 0000000110000000 1100000110000000 1100001100000000 0000011100000000 0000110000000000 0001100000000000 0011000000000000 0110001100000000 0111111100000000 0111111100000000	0111110000000000 1111111111100000 1111111111100000 1111111111100000 0000000110000000 0000001100000000 0000011100000000 0000111100000000 0001111100000000 0001110000000000 0001100000000000 0011100000000000 0011100000000000 0011000000000000 0011100000000000 0011110000000000	0001111100000000 0011111100000000 1111111000000000 1100111000000000 0000111110000000 0000111111000000 0001100011110000 0000000001110000 0000000000111000 0000000000111000 0000000000110000 0000000000110000 0000000011110000 0000000011100000 0000000110000000 0001111100000000	0011000000000000 1111000000000000 1111000000000000 1111000000000000 0111000000000000 0111000000000000 0111000000000000 0111000000000000 0011000000000000 0011000000000000 0011000000000000 0011000000000000 0011000000000000 0011000000000000 0011100000000000 0011100000000000	0000011100000000 0000011100000000 0000111100000000 0000111100000000 0001111100000000 0001111100000000 0011101110000000 0011000110000000 0011000110111000 0111111111111000 0111111111100000 1110000111110000 1100000111110000 0000000111110000 0000000111111100 0000000011111100
2	7	3	1	4

Test results for current version of captcha.dll using 100 downloaded captchas:

	00008.jpeg -> 00008
	01222.jpeg -> 01222
	01624.jpeg -> 01624
	02177.jpeg -> 02177
	03889.jpeg -> 03889
	04078.jpeg -> 04078
	06879.jpeg -> 06879
	06937.jpeg -> 06937
	08362.jpeg -> 08362
	10513.jpeg -> 10513
	10528.jpeg -> 10528
	10958.jpeg -> 10958
	13055.jpeg -> 13055
	16205.jpeg -> 16205
	16258.jpeg -> 16258
	17825.jpeg -> 17825
	18263.jpeg -> 18263
	19149.jpeg -> 19149
	19341.jpeg -> 19341
	20250.jpeg -> 20250
	21204.jpeg -> 21204
	21663.jpeg -> 21663
	22965.jpeg -> 22965
	24185.jpeg -> 24185
	24664.jpeg -> 24664
	25794.jpeg -> 25794
	26031.jpeg -> 26031
	27231.jpeg -> 27231
	27314.jpeg -> 27314
	27323.jpeg -> 27323
	28647.jpeg -> 28647
	31079.jpeg -> 31079
	31357.jpeg -> 31357
F	33956.jpeg -> 33958
	33971.jpeg -> 33971
	34550.jpeg -> 34550
	34753.jpeg -> 34753
	34983.jpeg -> 34983
	35456.jpeg -> 35456
	35597.jpeg -> 35597
	35998.jpeg -> 35998
	36845.jpeg -> 36845
	40272.jpeg -> 40272
	40813.jpeg -> 40813
	40949.jpeg -> 40949
	41600.jpeg -> 41600
	43161.jpeg -> 43161
	44631.jpeg -> 44631
	44689.jpeg -> 44689
	44898.jpeg -> 44898
	45455.jpeg -> 45455
	47255.jpeg -> 47255
	48984.jpeg -> 48984
	49333.jpeg -> 49333
	49840.jpeg -> 49840
	49984.jpeg -> 49984
	52493.jpeg -> 52493
	53535.jpeg -> 53535
	54001.jpeg -> 54001
	54003.jpeg -> 54003
	54768.jpeg -> 54768
	54950.jpeg -> 54950
	55319.jpeg -> 55319
	56024.jpeg -> 56024
	57225.jpeg -> 57225
	58820.jpeg -> 58820
	59031.jpeg -> 59031
	60244.jpeg -> 60244
	62986.jpeg -> 62986
	63239.jpeg -> 63239
	65479.jpeg -> 65479
	66362.jpeg -> 66362
	67201.jpeg -> 67201
	69675.jpeg -> 69675
	70296.jpeg -> 70296
	72494.jpeg -> 72494
	73160.jpeg -> 73160
	80813.jpeg -> 80813
	81145.jpeg -> 81145
	85139.jpeg -> 85139
	85624.jpeg -> 85624
	85943.jpeg -> 85943
	88056.jpeg -> 88056
	88187.jpeg -> 88187
	88190.jpeg -> 88190
	88514.jpeg -> 88514
	89146.jpeg -> 89146
	89192.jpeg -> 89192
	90697.jpeg -> 90697
	91462.jpeg -> 91462
	91746.jpeg -> 91746
	92897.jpeg -> 92897
	93735.jpeg -> 93735
F	96023.jpeg -> 96025
	97310.jpeg -> 97310
	97499.jpeg -> 97499
	97878.jpeg -> 97878
	98507.jpeg -> 98507
	98578.jpeg -> 98578
	99903.jpeg -> 99903

No Comments

You need to be logged in to be able to post comments