[千分悬赏]关于验证码识别中去除噪声的讨论（附单双噪点去除例子,详细注释可供学习）-已解决问题区-AUTOIT CN

tryhi 发表于 2013-4-19 16:33:19

[千分悬赏]关于验证码识别中去除噪声的讨论（附单双噪点去除例子,详细注释可供学习）

本帖最后由 tryhi 于 2013-4-19 21:24 编辑

其实之前开过这个贴，后来没人回复之后好像被关闭了http://www.autoitx.com/thread-37246-1-1.html
发出去的分收不回大家尽力拍砖。

注：图中的软件是用了疯子前辈的软件http://www.autoitx.com/forum.php?mod=viewthread&tid=32094
由于没学过验证码识别，所以提出的问题可能有点弱
图中红色(单点双点)、绿色(三点相连)的点是要去除的（所谓的红色绿色是我标出来，实际上图片已经二值化了，见代码），目前我只想到了红色点的去除，关于绿色点没有比较好的方案，红色的思路是这样的，搜索该点周围8个点，如果空白，则去除，如果周围只有1个点，则搜索这一个的周围8个点，发现它也只有一个，就两个都去除，这样就能实现去除红色点的效果，代码如下，可能效率、思路、代码都不是很好，关于三点相连或者抱团的点，应该怎么去除比较好？R的右腿那里有个很浅的红色圆圈那里三个点是不用去除的，怕有人误除，特指出来。

以下代码为去除单点双点的算法

;去除单双噪点算法例子
#include <Array.au3>
$string = "                                                                   0                      "&@CRLF& _
"                0 00                                                                         "&@CRLF& _
"                                     00                   0    0       0             "&@CRLF& _
" 0                                  0                            0          0    0 "&@CRLF& _
"                   0                      000                            0          "&@CRLF& _
"                               0                                           0             "&@CRLF& _
"       0                   0                                  0                      "&@CRLF& _
"                            0                0                00                      "&@CRLF& _
"                                                 0                               0       "&@CRLF& _
" 00    00       0                                                             0    "&@CRLF& _
"    00000000       0                                                          0       "&@CRLF& _
"       00          0             0 0                                  0000000    "&@CRLF& _
"       00                                           0             0 00000 000 "&@CRLF& _
"       00                      0                                           00    00 0 "&@CRLF& _
"       00                0                                                    00    00"&@CRLF& _
" 0 00                000000000                      0                00    00 "&@CRLF& _
"0    00 0          00 000             0000000000                00    00"&@CRLF& _
"    00          0    00 000          000    0    0       00    000 "&@CRLF& _
"    00                00    00          000 0          0 00 0 00"&@CRLF& _
"    00                   00 0 00          00 0                   00    00"&@CRLF& _
"    00       0       00    000 0       00                         00       00 "&@CRLF& _
"    00          0    00 00             0000                   000    00 "&@CRLF& _
"    00                   00 0000 0          00 0 0                00    00 "&@CRLF& _
" 0 00                   0000000    0       00 00                   00    00 "&@CRLF& _
"    00                   0000       0 0 00000000 0 0          000 00 "&@CRLF& _
"    00       0    0000 00                00 00 0             00    00"&@CRLF& _
"    00                   00 00                000 0                   00    0 "&@CRLF& _
"    00                   00 000             00 0                   00    00 "&@CRLF& _
"    00                   000000             00                         00    0"&@CRLF& _
"    00                   00 0000             00                      000 00"&@CRLF& _
"    00                   00 00             00                      0 00 000 "&@CRLF& _
" 0000                   00 00          0 00                   0 0000000 "&@CRLF& _
" 00000000       0    0000 000             00                      000       "&@CRLF& _
"                         00       0             000       0                   0"&@CRLF& _
"                0                      0    0 00       0 0                   "&@CRLF

Local $array_s;定义36x98的二维数组
$hang = StringRegExp($string,'\V+',3);===========此段代码为拆分进二维数组
For $i = 0 To 34
   $lie = StringSplit($hang[$i],'',2)
   For $j = 0 To 96
            $array_s[$i][$j] = $lie[$j]
   Next
Next;===================================================================
_quchuzaodian($array_s);主函数
_ArrayDisplay($array_s,'去除噪声之后');去除噪点之后的数组

Func _quchuzaodian(ByRef $array)
   $35 = UBound($array, 1) - 1
   $97 = UBound($array, 2) - 1
   For $i = 0 To $35;纵向历遍
            For $j = 0 To $97;横向历遍
                     If $array[$i][$j] = "0" Then;一路扫过去发现0停下搜索四周
                           $sizhou = _sizhou($array, $i, $j);搜索四周8个点，返回一个数组，为四周存在0的的数量
                           If $sizhou = 0 Then $array[$i][$j] = " ";发现四周不存在其他点，则去除该点（变为空格变量则去除）
                           If $sizhou = 1 Then;发现只存在一个点
                                    $sizhou2 = _sizhou($array, $sizhou, $sizhou);则搜索存在的这个点四周8个点
                                    If $sizhou2 = 1 Then;发现也只有一个点，也就是刚才那个点，则断定这是一个双点，可以去除这两个点
                                             $array[$i][$j] = " ";去除
                                             $array[$sizhou][$sizhou] = " ";去除
                                    EndIf
                           EndIf
                     EndIf
            Next
   Next
EndFunc ;==>_quchuzaodian

Func _sizhou($array, $1, $2, $front = '0')
   If Not IsArray($array) Then Return SetError(1)
   Local $a = UBound($array, 1) - 1, $b = UBound($array, 2) - 1;$a为数组一维下标-1，$b为数组二维下标-1
   Local $temp = ;定义一个数组用来存返回的结果0为四周存在点的数量,，若只有一个点，则，1为x坐标，2为y坐标
   If $1 > 0 And $2 > 0 Then;左上角的点1维2维均要减1，所以要保证都大于零
            If $array[$1 - 1][$2 - 1] = $front Then $temp = _sizhou_udf($temp, $1, $2, 1);左上角定义为1
EndIf
   If $1 > 0 Then;上面的点1维要减1，所以要保证大于零
            If $array[$1 - 1][$2 + 0] = $front Then $temp = _sizhou_udf($temp, $1, $2, 2);上面定义为2
EndIf
If $temp >= 2 Then Return $temp;当超出2个点时候再往下已经没有意义了
   If $1 > 0 And $2 < $b Then;左上角的点1维减1（保证大于0）,2维要加1（保证小于$b不能超界）
            If $array[$1 - 1][$2 + 1] = $front Then $temp = _sizhou_udf($temp, $1, $2, 3);右上角定义为3
   EndIf
If $temp >= 2 Then Return $temp;当超出2个点时候再往下已经没有意义了
   If $2 < $b Then
            If $array[$1 + 0][$2 + 1] = $front Then $temp = _sizhou_udf($temp, $1, $2, 4);右边定义为4
   EndIf
If $temp >= 2 Then Return $temp;当超出2个点时候再往下已经没有意义了
   If $1 < $a And $2 < $b Then
            If $array[$1 + 1][$2 + 1] = $front Then $temp = _sizhou_udf($temp, $1, $2, 5);右下角定义为5
   EndIf
If $temp >= 2 Then Return $temp;当超出2个点时候再往下已经没有意义了
   If $1 < $a Then
            If $array[$1 + 1][$2 + 0] = $front Then $temp = _sizhou_udf($temp, $1, $2, 6);下面定义为6
   EndIf
If $temp >= 2 Then Return $temp;当超出2个点时候再往下已经没有意义了
   If $1 < $a And $2 > 0 Then
            If $array[$1 + 1][$2 - 1] = $front Then $temp = _sizhou_udf($temp, $1, $2, 7);左下角定义为7
   EndIf
If $temp >= 2 Then Return $temp;当超出2个点时候再往下已经没有意义了
   If $2 > 0 Then
            If $array[$1 + 0][$2 - 1] = $front Then $temp = _sizhou_udf($temp, $1, $2, 8);左边定义为8
   EndIf
   Return $temp
EndFunc ;==>_sizhou

Func _sizhou_udf($array,$1,$2,$i)
   Switch $i;将、分别变成该点坐标，这样会导致只记录最后一个点的坐标，不过没事，当超出两个点时这两个变量就没用了
Case 1
                     $array = $1-1
                     $array = $2-1
            Case 2
                     $array = $1-1
                     $array = $2+0
            Case 3
                     $array =$1-1
                     $array = $2+1
            Case 4
                     $array = $1+0
                     $array = $2+1
            Case 5
                     $array = $1+1
                     $array = $2+1
            Case 6
                     $array = $1+1
                     $array = $2+0
            Case 7
                     $array = $1+1
                     $array = $2-1
            Case 8
                     $array = $1+0
                     $array = $2-1
   EndSwitch
   $array += 1
   Return $array
EndFunc

发现好多人看不懂我的表达，没办法将例子中的变量跟这个图联系起来，其实这个图只是表示一个二维数组
看到$string这个变量没有，把它拆进二维数组，一个子变量存一个字符，只有“0”或者一个空格，两种状态
你运行一下就会发现这个数组运行前跟运行后的变化
之前标记为红色的0变量全部成了空格变量
那有什么办法将标记为绿色(三个相连)的0变量也一并变成空格变量，就是一个这样的问题

Duvet 发表于 2013-4-19 16:33:20

本帖最后由 Duvet 于 2013-4-19 22:39 编辑

應該不是很難的問題,歸遞就可以解決了!?
使用歸遞的話效率就...

;~ Duvet
;~ 2013-04-19
#include <Array.au3>
Local $String, $Array, $Total = 4

$String = _
"0000000000000000000000000000000000000000000000000000000000000000000000100000000000000000000000000"&@CRLF& _
"0000000000000000001011000000000000000000000000000000000000000000000000000000000000000000000000000"&@CRLF& _
"0000000000000000000000000000000000000001100000000000000000000010000000100000000001000000000000000"&@CRLF& _
"0001000000000000000000000000000000000000100000000000000000000000000000010000000000000010000001000"&@CRLF& _
"0000000000000000000001000000000000000000000000011100000000000000000000000000000000100000000000000"&@CRLF& _
"0000000000000000000000000000000000100000000000000000000000000000000000000000000001000000000000000"&@CRLF& _
"0000000000100000000000000000000001000000000000000000000000000000000000001000000000000000000000000"&@CRLF& _
"0000000000000000000000000000000010000000000000000001000000000000000000011000000000000000000000000"&@CRLF& _
"0000000000000000000000000000000000000000000000000001000000000000000000000000000000000001000000000"&@CRLF& _
"0000011000000011000000000100000000000000000000000000000000000000000000000000000000000000001000000"&@CRLF& _
"0000000111111110000000000100000000000000000000000000000000000000000000000000000000000010000000000"&@CRLF& _
"0000000001100000000000000100000000000000010000010000000000000000000000000000000000001111111000000"&@CRLF& _
"0000000001100000000000000000000000000000000000000000000000100000000000000001000001111100001110000"&@CRLF& _
"0000000001100000000000000000000000010000000000000000000000000000000000000000000000011000000011010"&@CRLF& _
"0000000001100000000000000000010000000000000000000000000000000000000000000000000000001100000001100"&@CRLF& _
"0000100011000000000000000000011111111100000000000000000000000010000000000000000000011000000011000"&@CRLF& _
"1000000011000001000000000000001100001110000000000000000111111111100000000000000000011000000001100"&@CRLF& _
"0000000011000000000000010000001100000111000000000000011100000001000000010000000000011000000001110"&@CRLF& _
"0000000011000000000000000000001100000011000000000000001100100000100000000000000100011000100001100"&@CRLF& _
"0000000110000000000000000000001101000011000000000000001100010000000000000000000000011000000001100"&@CRLF& _
"0000000110000000000010000000001100000011100010000000001100000000000000000000000000011000000000110"&@CRLF& _
"0000000110000000000000010000001100000110000000000000001100100100000000000000000000011100000000110"&@CRLF& _
"0000000110000000000000000000001100001111010000000000001100000100010000000000000000001100000000110"&@CRLF& _
"0100000110000000000000000000001111111000000100000000001100000110000000000000000000001100000000110"&@CRLF& _
"0000000110000000000000000000001100110000000000010001000111111110001000100000000000001100100000110"&@CRLF& _
"0000000011000000000010000001001110001100000000000000000011000011000001000000000000000110000000011"&@CRLF& _
"0000000011000000000000000000000110001100000000000000000011100001000000000000000000000110000000010"&@CRLF& _
"0000000011000000000000000000000110100110000000000000000011000001000000000000000000000110000000110"&@CRLF& _
"0000000011000000000000000000000110011110000000000000000011000000000000000000000000000110000000100"&@CRLF& _
"0000000110000000000000000000000110000111100000000000000011000000000000000000000000010011000001100"&@CRLF& _
"0000000110000000000000000000000110000011000000000000000011000000000000000000000000100011000111000"&@CRLF& _
"0001100110000000000000000000000110000011000000000000010001100000000000000000000001000111111100000"&@CRLF& _
"0000011111111000000000010000001111000001110000000000000001100000000000000000000000001110000000000"&@CRLF& _
"0000000000000000000000000000011000000000001000000000000000111000000000100000000000000000000000100"&@CRLF& _
"0000000000000000001000000000000000000000000100000000100001100000000000100010000000000000000000000"&@CRLF

$Array = StringRegExp($String, "", 3)
$Array = _1DTo2D($Array, 97)
_CheckColor($Array)
_ArrayDisplay($Array)
Exit

Func _1DTo2D(ByRef $oArray, $endj)
Local $endi = Int(UBound($oArray) / $endj)
Local $nArray[$endi][$endj]
For $i = 0 To UBound($oArray) - 1
$nArray = Int($oArray[$i])
Next
Return $nArray
EndFunc

Func _CheckColor(ByRef $oArray)
For $i = 0 To UBound($oArray, 1) - 1
For $j = 0 To UBound($oArray, 2) - 1
If $oArray[$i][$j] Then
If __CheckColor($oArray, $i, $j, True) Then ContinueLoop
$oArray[$i][$j] = 0
EndIf
Next
Next
EndFunc

Func __CheckColor(ByRef $oArray, $i, $j, $start = False)
Local Static $k = 0, $up = 1, $down = 1, $left = 1, $right = 1
If $start Then
$k = 0
EndIf
If $k < $Total And $up And $left And $i And $j And $oArray[$i - 1][$j - 1] Then
$up = 1
$down = 0
$left = 1
$right = 0
$k += 1
__CheckColor($oArray, $i - 1, $j - 1)
EndIf
If $k < $Total And $up And $i And $oArray[$i - 1][$j] Then
$up = 1
$down = 0
$left = 1
$right = 1
$k += 1
__CheckColor($oArray, $i - 1, $j)
EndIf
If $k < $Total And $up And $right And $i And $j < UBound($oArray, 2) - 1 And $oArray[$i - 1][$j + 1] Then
$up = 1
$down = 0
$left = 0
$right = 1
$k += 1
__CheckColor($oArray, $i - 1, $j + 1)
EndIf
If $k < $Total And $left And $j And $oArray[$i][$j - 1] Then
$up = 1
$down = 1
$left = 1
$right = 0
$k += 1
__CheckColor($oArray, $i, $j - 1)
EndIf
If $k < $Total And $right And $j < UBound($oArray, 2) - 1 And $oArray[$i][$j + 1] Then
$up = 1
$down = 1
$left = 0
$right = 1
$k += 1
__CheckColor($oArray, $i, $j + 1)
EndIf
If $k < $Total And $down And $left And $i < UBound($oArray, 1) - 1 And $j And $oArray[$i + 1][$j - 1] Then
$up = 0
$down = 1
$left = 1
$right = 0
$k += 1
__CheckColor($oArray, $i + 1, $j - 1)
EndIf
If $k < $Total And $down And $i < UBound($oArray, 1) - 1 And $oArray[$i + 1][$j] Then
$up = 0
$down = 1
$left = 1
$right = 1
$k += 1
__CheckColor($oArray, $i + 1, $j)
EndIf
If $k < $Total And $down And $right And $i < UBound($oArray, 1) - 1 And $j < UBound($oArray, 2) - 1 And $oArray[$i + 1][$j + 1] Then
$up = 0
$down = 1
$left = 0
$right = 1
$k += 1
__CheckColor($oArray, $i + 1, $j + 1)
EndIf
$up = 1
$down = 1
$left = 1
$right = 1
Return $k = $Total
EndFunc

zldfsz 发表于 2013-4-19 19:40:20

不会，坐等高淫

haijie1223 发表于 2013-4-19 23:03:18

回复 2# Duvet

厉害！我一晚上了也没想到办法~

冷风无泪 发表于 2013-4-24 00:59:19

偶是不会的了，帮忙顶下帖子

pk196371 发表于 2013-4-24 11:32:30

不会这个、看别人高手怎么说。

seniors 发表于 2013-4-24 15:00:27

在学习做验证码，网上研究了好多去除噪点的算法
对这种噪点用中值滤波算法最好，只是我觉得验证码图片太小，滤波后字符会不清晰
送上滤波代码
我的图像数组是用的二维数组
Func Medianfilter($aArray)
Local $iwidth = UBound($aArray), $iheight = UBound($aArray, 2)
Local $a_return[$iwidth][$iheight]
Local $window, $min ,$temp, $k
For $y = 1 To $iheight - 2
For $x = 1 To $iwidth - 2
$k = 0
For $i = $x-1 To $x+1
For $j = $y -1 To $y +1
$window[$k] = $aArray[$i][$j]
$k +=1
Next
Next
For $m = 0 To 4
$min = $m
For $n = $m+1 To 8
If $window[$n] <$window[$min] Then
$min = $n
$temp = $window[$m]
$window[$m] = $window[$min]
$window[$min] = $temp
EndIf
Next
Next
$a_return[$x][$y] = $window
Next
Next
For $x = 0 To $iwidth -1
$a_return[$x] = $aArray[$x]
$a_return[$x][$iheight-1] = $aArray[$x][$iheight-1]
Next
For $y = 0 To $iheight -1
$a_return[$y] = $aArray[$y]
$a_return[$iwidth-1][$y] = $aArray[$iwidth -1][$y]
Next
Return $a_return
EndFunc

jhszs 发表于 2013-5-8 06:49:08

恩看了下特征码写到程序里有点大哈哈

页: [1]

AUTOIT CN's Archiver

[千分悬赏]关于验证码识别中去除噪声的讨论（附单双噪点去除例子,详细注释可供学习）