Matlab中的简单文本阅读器（OCR）

  • Z时代
  • 2024-01-10
  • 分类：问答

我正在尝试编写一个简单的程序来读取灰度的bmp图像。我有一组模式（“ I”以外的整个字母），我想匹配它们。我在Matlab中这样做有问题。

我到目前为止所得到的：

clear
clc
%set of patterns
BW1 = imread('alphabet.bmp');
patterns = bwlabel(~BW1);
patternStats = regionprops(patterns,'all');
patternNumber = size(patternStats);
imagePatternArray = cell(patternNumber);
%make cell array of pattern vectors
for i = 1:1:patternNumber
  imageMatrix = patternStats(i).Image;
  imageVector = imageMatrix(:);
  imagePatternArray{i} = imageVector;
end
%set of chars
BW2 = imread('text.bmp');
text = bwlabel(~BW2);
textStats = regionprops(text,'all');
letterNumber = size(textStats);
imageLetterArray = cell(letterNumber);
%make cell array of text vectors
for i = 1:1:letterNumber
  imageMatrix = textStats(i).Image;
  imageVector = imageMatrix(:);
  imageLetterArray{i} = imageVector;
end
%lookup table
charSet =['A','B','C','D','E','F','G','H','J','K','L','M','N','O','P','Q','R','S','T','U','V','W','X','Y','Z'];

现在，我想将模式向量与给定向量进行比较，但是它们具有不同的大小。

我怎样才能做到这一点？有一些特殊的比较功能吗？我应该在末尾加0，然后用计算距离pdist吗？

回答：

灵魂

clear
clc
%set of patterns
BW1 = imread('alphabet.bmp');
patterns = bwlabel(~BW1);
patternStats = regionprops(patterns,'all');
patternNumber = size(patternStats);
imagePatternArray = cell(patternNumber);
%make cell array of pattern Matrices
for i = 1:1:patternNumber
  imageMatrix = patternStats(i).Image;
  imageMatrix = imresize(imageMatrix, [25 20]);
  imagePatternArray{i} = imageMatrix;
end
%set of chars
BW2 = imread('kol_2.bmp');
BW2Gray = rgb2gray(BW2); %convert text to grayscale bmp - 0 OR 1
text = bwlabel(~BW2Gray);
textStats = regionprops(text,'all');
letterNumber = size(textStats);
imageLetterArray = cell(letterNumber);
%make cell array of text Matrices
for i = 1:1:letterNumber
  imageMatrix = textStats(i).Image;
  imageMatrix = imresize(imageMatrix, [25 20]);
  imageLetterArray{i} = imageMatrix;
end
%white spaces
whiteSpacesIndexes = [];
for i = 1:letterNumber - 1
  firstLetterBox = textStats(i).BoundingBox;
  positionFirstVector = [firstLetterBox(1), firstLetterBox(2)];
  secondLetterBox = textStats(i+1).BoundingBox;
  positionSecondVector = [secondLetterBox(1), secondLetterBox(2)];
  distanceVector = positionSecondVector - positionFirstVector;
  distance = norm(distanceVector)
  % if the distance between is bigger that letter width plus 1/3 of width, it is a whitespace
  bothLettersSize = firstLetterBox(3)  + secondLetterBox(3);
  noSpaceDistance =  bothLettersSize - bothLettersSize * 0.25; % - 25 per cent (heuristic value)
  if (distance > noSpaceDistance) %&& (abs(distanceVector(2)) > 1.0)
      whiteSpacesIndexes = [whiteSpacesIndexes, i + 1];
  end 
end
compareVector = size(patternNumber);
indexArray = size(letterNumber);
for i = 1:1:letterNumber
    for j = 1:1:patternNumber        
           correlationMatrix = normxcorr2(imagePatternArray{j},imageLetterArray{i});           
          compareVector(j) = max(abs(correlationMatrix(:)));
    end    
        [correlationMax,correlationIndex] = max(compareVector);
       indexArray(i) = correlationIndex;
end
%lookup table
charSet = ['A','B','C','D','E','F','G','H','J','K','L','M','N','O','P','Q','R','S','T','U','V','W','X','Y','Z'];
%outPut stream
outPut = size(letterNumber);
for i = 1:1:letterNumber
   outPut(i) = charSet(indexArray(i));
end
whiteSpaceNumber = size(whiteSpacesIndexes,2);
whiteSpacesIndexes = whiteSpacesIndexes + (0:numel(whiteSpacesIndexes)-1)
nFinal = numel(outPut)+numel(whiteSpacesIndexes );      %# New length of result with blanks
newstr = blanks(nFinal);                                %# Initialize the result as blanks
newstr(setdiff(1:nFinal,whiteSpacesIndexes )) = outPut

我很简单，有一些缺点，例如

  • 不读“我”

  • 只读取水平的文本

  • 空白空间检测应改进

以上是 Matlab中的简单文本阅读器（OCR） 的全部内容， 来源链接： utcz.com/qa/417591.html

回到顶部