public TessResult CleanEnglish(TessResult tessOcrResult)
{
if (!RemoveNoiseText.NotTooManyNoiseCharacters(tessOcrResult.tess_word3))
{
tessOcrResult.id = "-1";
}
else
{
tessOcrResult.tess_word3 = Regex.Replace(tessOcrResult.tess_word3, @"[^a-zA-Z0-9\s]", "");
}
return(tessOcrResult);
}
public static TessResult getDictionaryWord(TessResult tr, int dictionaryExactMatchStringLength)
{
_dictionaryExactMatchStringLength = dictionaryExactMatchStringLength;
if (tr.tess_word3.Contains("ouse"))
{
Console.WriteLine("debug");
}
try
{
if (tr.tess_word3 == null || tr.tess_word3.Length < _dictionaryExactMatchStringLength) //input is invalid
{
tr.id = "-1";
}
else
{
DictResult dictR;
if (!tr.tess_word3.Contains(" ") && !tr.tess_word3.Contains("\n")) // input is a single word
{
dictR = checkOneWord(tr.tess_word3, tr.front, tr.back);
}
else if (!tr.tess_word3.Contains("\n")) // input is a single line
{
dictR = checkOneLine(tr.tess_word3, tr.front, tr.back);
}
else // input is multi-lines
{
dictR = checkMultiLines(tr.tess_word3, tr.front, tr.back);
}
tr.dict_similarity = dictR.similarity;
tr.dict_word3 = dictR.text;
}
return(tr);
}
catch (Exception e)
{
Log.WriteLine("Check Dictionary: " + e.Message);
throw e;
}
}
public List<TessResult> Apply(string inputPath, string outputPath, string TesseractResultsJSONFileName)
{
string[] filePaths = Directory.GetFiles(inputPath, "*.png");
if (filePaths.Length == 0)
return null;
List<TessResult> tessOcrResultList = new List<TessResult>();
try
{
Log.WriteLine("Tessearct in progress...");
Page page;
for (int i = 0; i < filePaths.Length; i++)
{
string filename = Path.GetFileNameWithoutExtension(filePaths[i]);
String[] splitTokens = filename.Split('_');
if (splitTokens.Length != 11)
continue;
using (Image<Gray, Byte> image = new Image<Gray, byte>(filePaths[i]))
{
string text = "";
float conf = 0;
var img = Pix.LoadFromFile(filePaths[i]);
page = _engine.Process(img, PageSegMode.SingleBlock);
text = page.GetText();
conf = page.GetMeanConfidence();
page.Dispose();
TessResult tr = new TessResult();
if (text.Length > 0)
{
tr.id = splitTokens[0];
tr.tess_word3 = Regex.Replace(text, "\n\n", "");
tr.tess_raw3 = text;
tr.tess_cost3 = (-1 * conf) + 1;
tr.fileName = Path.GetFileName(filename);
tr.x = Convert.ToInt16(splitTokens[7]);
tr.y = Convert.ToInt16(splitTokens[8]);
tr.w = Convert.ToInt16(splitTokens[9]);
tr.h = Convert.ToInt16(splitTokens[10]);
tessOcrResultList.Add(tr);
}
}
}
Log.WriteLine("Tessearct finished");
return tessOcrResultList;
}
catch (Exception e)
{
Log.WriteLine(e.Message);
Log.WriteLine(e.Source);
Log.WriteLine(e.StackTrace);
throw;
}
}
public static TessResult getDictionaryWord(TessResult tr, int dictionaryExactMatchStringLength)
{
if (tr.tess_word3 == null || tr.tess_word3.Length < _minWordLengthForDictionaryComparison || tr.tess_word3.Length < dictionaryExactMatchStringLength)
{
tr.id = "-1";
return(tr);
}
List <string> dictionaryResultList = new List <string>();
List <string> tempInputFragments = new List <string>(tr.tess_word3.Split(' '));
if (tr.tess_word3.Split(' ').Length > 1)
{
tempInputFragments.Add(tr.tess_word3);
}
string[] InputFragments = new string[tempInputFragments.Count];
int len = 0;
foreach (string substr in tempInputFragments)
{
InputFragments[len++] = substr;
}
string FinalReplacement = "";
double maxSimilarity = 0;
double avg_similarity = 0;
int word_count = 0;
Dictionary <string, double> equalMaxSimilarDictWordList = new Dictionary <string, double>();
List <string> equalMinDistanceDictWordList = new List <string>();
string Replacement = "";
string combinedMatch = "";
string individualMatch = "";
List <string> sameMatch = new List <string>();
try
{
for (int k = 0; k < InputFragments.Length; k++)
{
equalMaxSimilarDictWordList.Clear();
maxSimilarity = 0;
if (InputFragments[k].Length < 2)
{
continue;
}
word_count++;
InputFragments[k] = InputFragments[k].ToLower();
int WordLength = InputFragments[k].Length;
if (WordLength == dictionaryExactMatchStringLength)
{
maxSimilarity = findSimilarDictionaryWord(InputFragments[k], maxSimilarity, 0, equalMinDistanceDictWordList);
if (maxSimilarity != 1)
{
maxSimilarity = 0;
}
else
{
combinedMatch += equalMinDistanceDictWordList[0];
}
}
else
{
for (int m = 0; m < _maxWordLength; m++)
{
maxSimilarity = findSimilarDictionaryWord(InputFragments[k], maxSimilarity, m, equalMinDistanceDictWordList);
}
}
avg_similarity += maxSimilarity;
if (maxSimilarity > 0.33)
{
combinedMatch += equalMinDistanceDictWordList[0];
}
//if (maxSimilarity < 0.33) //dictionary word not found (most similar is 1) hill vs hall = 0.333333
// Replacement = InputFragments[k];
//else
/*{
* if (k < InputFragments.Length - 1)
* {
* combinedMatch += tempReplacement;
* combinedMatch += " ";
* }
* }
*
* if (k < InputFragments.Length - 1)
* {
* Replacement = "";
* foreach (var item in equalMaxSimilarDictWordList)
* {
* if (item.Value == maxSimilarity)
* {
* Replacement = item.Key;
* break;
* }
* }
* combinedMatch += Replacement + " ";
* }*/
}
equalMaxSimilarDictWordList.Add(combinedMatch, 1.1);
double maxscore = -1;
string maxstr = "";
Dictionary <string, int> matchVal = new Dictionary <string, int>();
foreach (var item in equalMaxSimilarDictWordList)
{
int score = NeedlemanWunsch.findSimScore(item.Key, tr.tess_word3);
matchVal.Add(item.Key, score);
if (score > maxscore)
{
maxscore = score;
maxstr = item.Key;
}
}
foreach (var sameitem in matchVal)
{
if (maxscore == sameitem.Value)
{
sameMatch.Add(sameitem.Key);
}
}
foreach (var item in equalMinDistanceDictWordList)
{
int score = NeedlemanWunsch.findSimScore(item, tr.tess_word3);
matchVal.Add(item, score);
if (score > maxscore)
{
maxscore = score;
maxstr = item;
}
}
foreach (var sameitem in matchVal)
{
if (maxscore == sameitem.Value)
{
sameMatch.Add(sameitem.Key);
}
}
// Replacement += equalMaxSimilarDictWordList[0]; // get the first dictionary word
Replacement = maxstr;
FinalReplacement = Replacement;
maxSimilarity = maxscore;
}
catch (Exception e)
{
Log.WriteLine("Check Dictionary: " + e.Message);
throw e;
}
tr.dict_word3 = FinalReplacement.Trim();
sameMatch = sameMatch.Distinct().ToList();
string sameMatches = "";
// sameMatches = String.Join(" , ", sameMatch);
foreach (string str in sameMatch)
{
sameMatches += str + " , ";
}
if (sameMatches.Length > 0)
{
sameMatches = sameMatches.Remove(sameMatches.Length - 1);
}
tr.sameMatches = sameMatches;
if (word_count == 0)
{
tr.dict_similarity = 0;
}
else
{
tr.dict_similarity = maxSimilarity;
}
// tr.dict_similarity = avg_similarity / Convert.ToDouble(word_count);
return(tr);
}
public static TessResult getDictionaryWord(TessResult tr, int dictionaryExactMatchStringLength)
{
_dictionaryExactMatchStringLength = dictionaryExactMatchStringLength;
if (tr.tess_word3.Contains("ouse"))
Console.WriteLine("debug");
try
{
if (tr.tess_word3 == null || tr.tess_word3.Length < _dictionaryExactMatchStringLength) //input is invalid
{
tr.id = "-1";
}
else
{
DictResult dictR;
if (!tr.tess_word3.Contains(" ") && !tr.tess_word3.Contains("\n")) // input is a single word
dictR = checkOneWord(tr.tess_word3, tr.front, tr.back);
else if (!tr.tess_word3.Contains("\n")) // input is a single line
dictR = checkOneLine(tr.tess_word3, tr.front, tr.back);
else // input is multi-lines
dictR = checkMultiLines(tr.tess_word3, tr.front, tr.back);
tr.dict_similarity = dictR.similarity;
tr.dict_word3 = dictR.text;
}
return tr;
}
catch (Exception e)
{
Log.WriteLine("Check Dictionary: " + e.Message);
throw e;
}
}
public TessResult CleanEnglish(TessResult tessOcrResult)
{
if (!RemoveNoiseText.NotTooManyNoiseCharacters(tessOcrResult.tess_word3))
tessOcrResult.id = "-1";
else
tessOcrResult.tess_word3 = Regex.Replace(tessOcrResult.tess_word3, @"[^a-zA-Z0-9\s]", "");
return tessOcrResult;
}
public TessResult CleanChinese(TessResult tessOcrResult)
{
tessOcrResult.tess_word3 = Regex.Replace(tessOcrResult.tess_word3, @"[^\u4E00-\u9FFF\u6300-\u77FF\u7800-\u8CFF\u8D00-\u9FFF]", "");
return tessOcrResult;
}
public TessResult CleanChinese(TessResult tessOcrResult)
{
tessOcrResult.tess_word3 = Regex.Replace(tessOcrResult.tess_word3, @"[^\u4E00-\u9FFF\u6300-\u77FF\u7800-\u8CFF\u8D00-\u9FFF]", "");
return(tessOcrResult);
}
public List <TessResult> Apply(string inputPath, string outputPath, string TesseractResultsJSONFileName)
{
string[] filePaths = Directory.GetFiles(inputPath, "*.png");
if (filePaths.Length == 0)
{
return(null);
}
List <TessResult> tessOcrResultList = new List <TessResult>();
try
{
Log.WriteLine("Tessearct in progress...");
Page page;
for (int i = 0; i < filePaths.Length; i++)
{
string filename = Path.GetFileNameWithoutExtension(filePaths[i]);
String[] splitTokens = filename.Split('_');
if (splitTokens.Length != 11)
{
continue;
}
using (Image <Gray, Byte> image = new Image <Gray, byte>(filePaths[i]))
{
string text = "";
float conf = 0;
var img = Pix.LoadFromFile(filePaths[i]);
page = _engine.Process(img, PageSegMode.SingleBlock);
text = page.GetText();
conf = page.GetMeanConfidence();
page.Dispose();
TessResult tr = new TessResult();
if (text.Length > 0)
{
tr.id = splitTokens[0];
tr.tess_word3 = Regex.Replace(text, "\n\n", "");
tr.tess_raw3 = text;
tr.tess_cost3 = (-1 * conf) + 1;
tr.fileName = Path.GetFileName(filename);
tr.x = Convert.ToInt16(splitTokens[7]);
tr.y = Convert.ToInt16(splitTokens[8]);
tr.w = Convert.ToInt16(splitTokens[9]);
tr.h = Convert.ToInt16(splitTokens[10]);
tessOcrResultList.Add(tr);
}
}
}
Log.WriteLine("Tessearct finished");
return(tessOcrResultList);
}
catch (Exception e)
{
Log.WriteLine(e.Message);
Log.WriteLine(e.Source);
Log.WriteLine(e.StackTrace);
throw;
}
}
public static TessResult getDictionaryWord(TessResult tr, int dictionaryExactMatchStringLength)
{
if (tr.tess_word3 == null || tr.tess_word3.Length < _minWordLengthForDictionaryComparison || tr.tess_word3.Length < dictionaryExactMatchStringLength)
{
tr.id ="-1";
return tr;
}
List<string> dictionaryResultList = new List<string>();
List<string> tempInputFragments = new List<string>(tr.tess_word3.Split(' '));
if (tr.tess_word3.Split(' ').Length > 1)
tempInputFragments.Add(tr.tess_word3);
string[] InputFragments = new string[tempInputFragments.Count];
int len = 0;
foreach (string substr in tempInputFragments)
{
InputFragments[len++] = substr;
}
string FinalReplacement = "";
double maxSimilarity = 0;
double avg_similarity = 0;
int word_count = 0;
Dictionary<string, double> equalMaxSimilarDictWordList = new Dictionary<string, double>();
List<string> equalMinDistanceDictWordList = new List<string>();
string Replacement = "";
string combinedMatch = "";
string individualMatch = "";
List<string> sameMatch = new List<string>();
try
{
for (int k = 0; k < InputFragments.Length; k++)
{
equalMaxSimilarDictWordList.Clear();
maxSimilarity = 0;
if (InputFragments[k].Length < 2)
continue;
word_count++;
InputFragments[k] = InputFragments[k].ToLower();
int WordLength = InputFragments[k].Length;
if(WordLength==dictionaryExactMatchStringLength)
{
maxSimilarity = findSimilarDictionaryWord(InputFragments[k], maxSimilarity, 0, equalMinDistanceDictWordList);
if (maxSimilarity != 1)
{
maxSimilarity = 0;
}
else
combinedMatch+=equalMinDistanceDictWordList[0];
}
else
for (int m = 0; m < _maxWordLength; m++)
{
maxSimilarity = findSimilarDictionaryWord(InputFragments[k], maxSimilarity, m, equalMinDistanceDictWordList);
}
avg_similarity += maxSimilarity;
if (maxSimilarity > 0.33)
combinedMatch += equalMinDistanceDictWordList[0];
//if (maxSimilarity < 0.33) //dictionary word not found (most similar is 1) hill vs hall = 0.333333
// Replacement = InputFragments[k];
//else
/*{
if (k < InputFragments.Length - 1)
{
combinedMatch += tempReplacement;
combinedMatch += " ";
}
}
if (k < InputFragments.Length - 1)
{
Replacement = "";
foreach (var item in equalMaxSimilarDictWordList)
{
if (item.Value == maxSimilarity)
{
Replacement = item.Key;
break;
}
}
combinedMatch += Replacement + " ";
}*/
}
equalMaxSimilarDictWordList.Add(combinedMatch, 1.1);
double maxscore = -1;
string maxstr = "";
Dictionary<string, int> matchVal = new Dictionary<string, int>();
foreach (var item in equalMaxSimilarDictWordList)
{
int score = NeedlemanWunsch.findSimScore(item.Key, tr.tess_word3);
matchVal.Add(item.Key, score);
if (score > maxscore)
{
maxscore = score;
maxstr = item.Key;
}
}
foreach (var sameitem in matchVal)
{
if (maxscore == sameitem.Value)
sameMatch.Add(sameitem.Key);
}
foreach (var item in equalMinDistanceDictWordList)
{
int score = NeedlemanWunsch.findSimScore(item, tr.tess_word3);
matchVal.Add(item, score);
if (score > maxscore)
{
maxscore = score;
maxstr = item;
}
}
foreach (var sameitem in matchVal)
{
if (maxscore == sameitem.Value)
sameMatch.Add(sameitem.Key);
}
// Replacement += equalMaxSimilarDictWordList[0]; // get the first dictionary word
Replacement = maxstr;
FinalReplacement = Replacement;
maxSimilarity = maxscore;
}
catch (Exception e)
{
Log.WriteLine("Check Dictionary: " + e.Message);
throw e;
}
tr.dict_word3 = FinalReplacement.Trim();
sameMatch = sameMatch.Distinct().ToList();
string sameMatches = "";
// sameMatches = String.Join(" , ", sameMatch);
foreach (string str in sameMatch)
sameMatches += str + " , ";
if (sameMatches.Length > 0)
sameMatches = sameMatches.Remove(sameMatches.Length - 1);
tr.sameMatches = sameMatches;
if (word_count == 0)
tr.dict_similarity = 0;
else
tr.dict_similarity = maxSimilarity;
// tr.dict_similarity = avg_similarity / Convert.ToDouble(word_count);
return tr;
}
