private string OcrViaTesseract(Bitmap bitmap, int index)
{
if (bitmap == null)
return string.Empty;
if (_ocrFixEngine == null)
comboBoxDictionaries_SelectedIndexChanged(null, null);
const int badWords = 0;
string textWithOutFixes;
if (_tesseractAsyncStrings != null && !string.IsNullOrEmpty(_tesseractAsyncStrings[index]))
{
textWithOutFixes = _tesseractAsyncStrings[index];
}
else
{
if (_tesseractAsyncIndex <= index)
_tesseractAsyncIndex = index + 10;
textWithOutFixes = Tesseract3DoOcrViaExe(bitmap, _languageId, "-psm 6"); // 6 = Assume a single uniform block of text.
}
if ((!textWithOutFixes.Contains(Environment.NewLine) || Utilities.CountTagInText(textWithOutFixes, '\n') > 2)
&& (textWithOutFixes.Length < 17 || bitmap.Height < 50))
{
string psm = Tesseract3DoOcrViaExe(bitmap, _languageId, "-psm 7"); // 7 = Treat the image as a single text line.
if (textWithOutFixes != psm)
{
if (string.IsNullOrWhiteSpace(textWithOutFixes))
{
textWithOutFixes = psm;
}
else if (psm.Length > textWithOutFixes.Length)
{
if (!psm.Contains('9') && textWithOutFixes.Contains('9') ||
!psm.Contains('6') && textWithOutFixes.Contains('6') ||
!psm.Contains('5') && textWithOutFixes.Contains('5') ||
!psm.Contains('3') && textWithOutFixes.Contains('3') ||
!psm.Contains('1') && textWithOutFixes.Contains('1') ||
!psm.Contains('$') && textWithOutFixes.Contains('$') ||
!psm.Contains('•') && textWithOutFixes.Contains('•') ||
!psm.Contains('Y') && textWithOutFixes.Contains('Y') ||
!psm.Contains('\'') && textWithOutFixes.Contains('\'') ||
!psm.Contains('€') && textWithOutFixes.Contains('€'))
{
textWithOutFixes = psm;
}
else if (_ocrFixEngine != null && !psm.Contains('$') && !psm.Contains('•') && !psm.Contains('€'))
{
int correctWordsNoFixes;
int wordsNotFoundNoFixes = _ocrFixEngine.CountUnknownWordsViaDictionary(textWithOutFixes, out correctWordsNoFixes);
int correctWordsPsm7;
int wordsNotFoundPsm7 = _ocrFixEngine.CountUnknownWordsViaDictionary(psm, out correctWordsPsm7);
if (wordsNotFoundPsm7 <= wordsNotFoundNoFixes && correctWordsPsm7 > correctWordsNoFixes)
{
textWithOutFixes = psm;
}
}
}
else if (psm.Length == textWithOutFixes.Length &&
(!psm.Contains('0') && textWithOutFixes.Contains('0') || // these chars are often mistaken
!psm.Contains('9') && textWithOutFixes.Contains('9') ||
!psm.Contains('8') && textWithOutFixes.Contains('8') ||
!psm.Contains('5') && textWithOutFixes.Contains('5') ||
!psm.Contains('3') && textWithOutFixes.Contains('3') ||
!psm.Contains('1') && textWithOutFixes.Contains('1') ||
!psm.Contains('$') && textWithOutFixes.Contains('$') ||
!psm.Contains('€') && textWithOutFixes.Contains('€') ||
!psm.Contains('•') && textWithOutFixes.Contains('•') ||
!psm.Contains('Y') && textWithOutFixes.Contains('Y') ||
!psm.Contains('\'') && textWithOutFixes.Contains('\'') ||
!psm.Contains('/') && textWithOutFixes.Contains('/') ||
!psm.Contains('(') && textWithOutFixes.Contains('(') ||
!psm.Contains(')') && textWithOutFixes.Contains(')') ||
!psm.Contains('_') && textWithOutFixes.Contains('_')))
{
textWithOutFixes = psm;
}
else if (psm.Length == textWithOutFixes.Length && psm.EndsWith('.') && !textWithOutFixes.EndsWith('.'))
{
textWithOutFixes = psm;
}
}
}
if (!checkBoxTesseractItalicsOn.Checked)
textWithOutFixes = HtmlUtil.RemoveOpenCloseTags(textWithOutFixes, HtmlUtil.TagItalic);
// Sometimes Tesseract has problems with small fonts - it helps to make the image larger
if (HtmlUtil.RemoveOpenCloseTags(textWithOutFixes, HtmlUtil.TagItalic).Replace("@", string.Empty).Replace("%", string.Empty).Replace("|", string.Empty).Trim().Length < 3
|| Utilities.CountTagInText(textWithOutFixes, '\n') > 2)
{
string rs = TesseractResizeAndRetry(bitmap);
textWithOutFixes = rs;
if (!checkBoxTesseractItalicsOn.Checked)
textWithOutFixes = HtmlUtil.RemoveOpenCloseTags(textWithOutFixes, HtmlUtil.TagItalic);
}
// fix italics
textWithOutFixes = FixItalics(textWithOutFixes);
int numberOfWords = textWithOutFixes.Split((" " + Environment.NewLine).ToCharArray(), StringSplitOptions.RemoveEmptyEntries).Length;
string line = textWithOutFixes.Trim();
if (_ocrFixEngine.IsDictionaryLoaded)
{
if (checkBoxAutoFixCommonErrors.Checked)
line = _ocrFixEngine.FixOcrErrors(line, index, _lastLine, true, GetAutoGuessLevel());
int correctWords;
int wordsNotFound = _ocrFixEngine.CountUnknownWordsViaDictionary(line, out correctWords);
int oldCorrectWords = correctWords;
if (wordsNotFound > 0 || correctWords == 0)
{
List<string> oldUnkownWords = new List<string>();
oldUnkownWords.AddRange(_ocrFixEngine.UnknownWordsFound);
_ocrFixEngine.UnknownWordsFound.Clear();
string newUnfixedText = TesseractResizeAndRetry(bitmap);
string newText = _ocrFixEngine.FixOcrErrors(newUnfixedText, index, _lastLine, true, GetAutoGuessLevel());
int newWordsNotFound = _ocrFixEngine.CountUnknownWordsViaDictionary(newText, out correctWords);
if (wordsNotFound == 1 && newWordsNotFound == 1 && newUnfixedText.EndsWith("!!") && textWithOutFixes.EndsWith('u') && newText.Length > 1)
{
_ocrFixEngine.UnknownWordsFound.Clear();
newText = textWithOutFixes.Substring(0, textWithOutFixes.Length - 1) + "!!";
newWordsNotFound = _ocrFixEngine.CountUnknownWordsViaDictionary(newText, out correctWords);
}
else if (correctWords >= oldCorrectWords &&
(!newText.Contains('9') || textWithOutFixes.Contains('9')) &&
(!newText.Replace("</i>", string.Empty).Contains('/') || textWithOutFixes.Replace("</i>", string.Empty).Contains('/')) &&
!string.IsNullOrWhiteSpace(newUnfixedText) &&
newWordsNotFound < wordsNotFound || (newWordsNotFound == wordsNotFound && newText.EndsWith('!') && textWithOutFixes.EndsWith('l')))
{
wordsNotFound = newWordsNotFound;
if (textWithOutFixes.Length > 3 && textWithOutFixes.EndsWith("...") && !newText.EndsWith('.') && !newText.EndsWith(',') && !newText.EndsWith('!') &&
!newText.EndsWith('?') && !newText.EndsWith("</i>"))
newText = newText.TrimEnd() + "...";
else if (textWithOutFixes.Length > 0 && textWithOutFixes.EndsWith('.') && !newText.EndsWith('.') && !newText.EndsWith(',') && !newText.EndsWith('!') &&
!newText.EndsWith('?') && !newText.EndsWith("</i>"))
newText = newText.TrimEnd() + ".";
else if (textWithOutFixes.Length > 0 && textWithOutFixes.EndsWith('?') && !newText.EndsWith('.') && !newText.EndsWith(',') && !newText.EndsWith('!') &&
!newText.EndsWith('?') && !newText.EndsWith("</i>"))
newText = newText.TrimEnd() + "?";
textWithOutFixes = newUnfixedText;
line = FixItalics(newText);
}
else if (correctWords > oldCorrectWords + 1 || (correctWords > oldCorrectWords && !textWithOutFixes.Contains(' ')))
{
wordsNotFound = newWordsNotFound;
textWithOutFixes = newUnfixedText;
line = newText;
}
else
{
_ocrFixEngine.UnknownWordsFound.Clear();
_ocrFixEngine.UnknownWordsFound.AddRange(oldUnkownWords);
}
}
if (wordsNotFound > 0 || correctWords == 0 || textWithOutFixes != null && textWithOutFixes.Replace("~", string.Empty).Trim().Length < 2)
{
if (_bluRaySubtitles != null && !line.Contains("<i>"))
{
_ocrFixEngine.AutoGuessesUsed.Clear();
_ocrFixEngine.UnknownWordsFound.Clear();
// which is best - normal image or one color image?
var nbmp = new NikseBitmap(bitmap);
nbmp.MakeOneColor(Color.White);
Bitmap oneColorBitmap = nbmp.GetBitmap();
string oneColorText = Tesseract3DoOcrViaExe(oneColorBitmap, _languageId, "-psm 6"); // 6 = Assume a single uniform block of text.
oneColorBitmap.Dispose();
nbmp = null;
if (oneColorText.Length > 1 &&
!oneColorText.Contains("CD") &&
(!oneColorText.Contains('0') || line.Contains('0')) &&
(!oneColorText.Contains('2') || line.Contains('2')) &&
(!oneColorText.Contains('3') || line.Contains('4')) &&
(!oneColorText.Contains('5') || line.Contains('5')) &&
(!oneColorText.Contains('9') || line.Contains('9')) &&
(!oneColorText.Contains('•') || line.Contains('•')) &&
(!oneColorText.Contains(')') || line.Contains(')')) &&
Utilities.CountTagInText(oneColorText, '(') < 2 && Utilities.CountTagInText(oneColorText, ')') < 2 &&
Utilities.GetNumberOfLines(oneColorText) < 4)
{
int modiCorrectWords;
int modiWordsNotFound = _ocrFixEngine.CountUnknownWordsViaDictionary(oneColorText, out modiCorrectWords);
string modiTextOcrFixed = oneColorText;
if (checkBoxAutoFixCommonErrors.Checked)
modiTextOcrFixed = _ocrFixEngine.FixOcrErrors(oneColorText, index, _lastLine, false, GetAutoGuessLevel());
int modiOcrCorrectedCorrectWords;
int modiOcrCorrectedWordsNotFound = _ocrFixEngine.CountUnknownWordsViaDictionary(modiTextOcrFixed, out modiOcrCorrectedCorrectWords);
if (modiOcrCorrectedWordsNotFound <= modiWordsNotFound)
{
oneColorText = modiTextOcrFixed;
modiWordsNotFound = modiOcrCorrectedWordsNotFound;
modiCorrectWords = modiOcrCorrectedCorrectWords;
}
if (modiWordsNotFound < wordsNotFound || (textWithOutFixes.Length == 1 && modiWordsNotFound == 0))
{
line = FixItalics(oneColorText); // use one-color text
wordsNotFound = modiWordsNotFound;
correctWords = modiCorrectWords;
if (checkBoxAutoFixCommonErrors.Checked)
line = _ocrFixEngine.FixOcrErrors(line, index, _lastLine, true, GetAutoGuessLevel());
}
else if (wordsNotFound == modiWordsNotFound && oneColorText.EndsWith('!') && (line.EndsWith('l') || line.EndsWith('fl')))
{
line = FixItalics(oneColorText);
wordsNotFound = modiWordsNotFound;
correctWords = modiCorrectWords;
if (checkBoxAutoFixCommonErrors.Checked)
line = _ocrFixEngine.FixOcrErrors(line, index, _lastLine, true, GetAutoGuessLevel());
}
}
}
}
if (checkBoxTesseractItalicsOn.Checked)
{
if (line.Contains("<i>") || wordsNotFound > 0 || correctWords == 0 || textWithOutFixes != null && textWithOutFixes.Replace("~", string.Empty).Trim().Length < 2)
{
_ocrFixEngine.AutoGuessesUsed.Clear();
_ocrFixEngine.UnknownWordsFound.Clear();
// which is best - normal image or de-italic'ed? We find out here
var unItalicedBmp = UnItalic(bitmap, _unItalicFactor);
string unItalicText = Tesseract3DoOcrViaExe(unItalicedBmp, _languageId, "-psm 6"); // 6 = Assume a single uniform block of text.
unItalicedBmp.Dispose();
if (unItalicText.Length > 1)
{
int modiCorrectWords;
int modiWordsNotFound = _ocrFixEngine.CountUnknownWordsViaDictionary(unItalicText, out modiCorrectWords);
string modiTextOcrFixed = unItalicText;
if (checkBoxAutoFixCommonErrors.Checked)
modiTextOcrFixed = _ocrFixEngine.FixOcrErrors(unItalicText, index, _lastLine, false, GetAutoGuessLevel());
int modiOcrCorrectedCorrectWords;
int modiOcrCorrectedWordsNotFound = _ocrFixEngine.CountUnknownWordsViaDictionary(modiTextOcrFixed, out modiOcrCorrectedCorrectWords);
if (modiOcrCorrectedWordsNotFound <= modiWordsNotFound)
{
unItalicText = modiTextOcrFixed;
modiWordsNotFound = modiOcrCorrectedWordsNotFound;
modiCorrectWords = modiOcrCorrectedCorrectWords;
}
bool ok = modiWordsNotFound < wordsNotFound || (textWithOutFixes.Length == 1 && modiWordsNotFound == 0);
if (!ok)
ok = wordsNotFound == modiWordsNotFound && unItalicText.EndsWith('!') && (line.EndsWith('l') || line.EndsWith('fl'));
if (!ok)
ok = wordsNotFound == modiWordsNotFound && line.StartsWith("<i>") && line.EndsWith("</i>");
if (ok && Utilities.CountTagInText(unItalicText, '/') > Utilities.CountTagInText(line, '/') + 1)
ok = false;
if (ok && Utilities.CountTagInText(unItalicText, '\\') > Utilities.CountTagInText(line, '\\'))
ok = false;
if (ok && Utilities.CountTagInText(unItalicText, ')') > Utilities.CountTagInText(line, ')') + 1)
ok = false;
if (ok && Utilities.CountTagInText(unItalicText, '(') > Utilities.CountTagInText(line, '(') + 1)
ok = false;
if (ok && Utilities.CountTagInText(unItalicText, '$') > Utilities.CountTagInText(line, '$') + 1)
ok = false;
if (ok && Utilities.CountTagInText(unItalicText, '€') > Utilities.CountTagInText(line, '€') + 1)
ok = false;
if (ok && Utilities.CountTagInText(unItalicText, '•') > Utilities.CountTagInText(line, '•'))
ok = false;
if (ok)
{
wordsNotFound = modiWordsNotFound;
correctWords = modiCorrectWords;
line = HtmlUtil.RemoveOpenCloseTags(line, HtmlUtil.TagItalic).Trim();
if (line.Length > 7 && unItalicText.Length > 7 && unItalicText.StartsWith("I ") &&
line.StartsWith(unItalicText.Remove(0, 2).Substring(0, 4)))
unItalicText = unItalicText.Remove(0, 2);
if (checkBoxTesseractMusicOn.Checked)
{
if ((line.StartsWith("J' ") || line.StartsWith("J“ ") || line.StartsWith("J* ") || line.StartsWith("♪ ")) && unItalicText.Length > 3 && HtmlUtil.RemoveOpenCloseTags(unItalicText, HtmlUtil.TagItalic).Substring(1, 2) == "' ")
{
unItalicText = "♪ " + unItalicText.Remove(0, 2).TrimStart();
}
if ((line.StartsWith("J' ") || line.StartsWith("J“ ") || line.StartsWith("J* ") || line.StartsWith("♪ ")) && unItalicText.Length > 3 && HtmlUtil.RemoveOpenCloseTags(unItalicText, HtmlUtil.TagItalic)[1] == ' ')
{
bool ita = unItalicText.StartsWith("<i>") && unItalicText.EndsWith("</i>");
unItalicText = HtmlUtil.RemoveHtmlTags(unItalicText);
unItalicText = "♪ " + unItalicText.Remove(0, 2).TrimStart();
if (ita)
unItalicText = "<i>" + unItalicText + "</i>";
}
if ((line.StartsWith("J' ") || line.StartsWith("J“ ") || line.StartsWith("J* ") || line.StartsWith("♪ ")) && unItalicText.Length > 3 && HtmlUtil.RemoveOpenCloseTags(unItalicText, HtmlUtil.TagItalic)[2] == ' ')
{
bool ita = unItalicText.StartsWith("<i>") && unItalicText.EndsWith("</i>");
unItalicText = HtmlUtil.RemoveHtmlTags(unItalicText);
unItalicText = "♪ " + unItalicText.Remove(0, 2).TrimStart();
if (ita)
unItalicText = "<i>" + unItalicText + "</i>";
}
if (unItalicText.StartsWith("J'") && (line.StartsWith('♪') || textWithOutFixes.StartsWith('♪') || textWithOutFixes.StartsWith("<i>♪") || unItalicText.EndsWith('♪')))
{
bool ita = unItalicText.StartsWith("<i>") && unItalicText.EndsWith("</i>");
unItalicText = HtmlUtil.RemoveHtmlTags(unItalicText);
unItalicText = "♪ " + unItalicText.Remove(0, 2).TrimStart();
if (ita)
unItalicText = "<i>" + unItalicText + "</i>";
}
if ((line.StartsWith("J` ") || line.StartsWith("J“ ") || line.StartsWith("J' ") || line.StartsWith("J* ")) && unItalicText.StartsWith("S "))
{
bool ita = unItalicText.StartsWith("<i>") && unItalicText.EndsWith("</i>");
unItalicText = HtmlUtil.RemoveHtmlTags(unItalicText);
unItalicText = "♪ " + unItalicText.Remove(0, 2).TrimStart();
if (ita)
unItalicText = "<i>" + unItalicText + "</i>";
}
if ((line.StartsWith("J` ") || line.StartsWith("J“ ") || line.StartsWith("J' ") || line.StartsWith("J* ")) && unItalicText.StartsWith("<i>S</i> "))
{
bool ita = unItalicText.StartsWith("<i>") && unItalicText.EndsWith("</i>");
unItalicText = HtmlUtil.RemoveHtmlTags(unItalicText);
unItalicText = "♪ " + unItalicText.Remove(0, 8).TrimStart();
if (ita)
unItalicText = "<i>" + unItalicText + "</i>";
}
if (unItalicText.StartsWith(";'") && (line.StartsWith('♪') || textWithOutFixes.StartsWith('♪') || textWithOutFixes.StartsWith("<i>♪") || unItalicText.EndsWith('♪')))
{
bool ita = unItalicText.StartsWith("<i>") && unItalicText.EndsWith("</i>");
unItalicText = HtmlUtil.RemoveHtmlTags(unItalicText);
unItalicText = "♪ " + unItalicText.Remove(0, 2).TrimStart();
if (ita)
unItalicText = "<i>" + unItalicText + "</i>";
}
if (unItalicText.StartsWith(",{*") && (line.StartsWith('♪') || textWithOutFixes.StartsWith('♪') || textWithOutFixes.StartsWith("<i>♪") || unItalicText.EndsWith('♪')))
{
bool ita = unItalicText.StartsWith("<i>") && unItalicText.EndsWith("</i>");
unItalicText = HtmlUtil.RemoveHtmlTags(unItalicText);
unItalicText = "♪ " + unItalicText.Remove(0, 3).TrimStart();
if (ita)
unItalicText = "<i>" + unItalicText + "</i>";
}
if (unItalicText.EndsWith("J'") && (line.EndsWith('♪') || textWithOutFixes.EndsWith('♪') || textWithOutFixes.EndsWith("♪</i>") || unItalicText.StartsWith('♪')))
{
bool ita = unItalicText.StartsWith("<i>") && unItalicText.EndsWith("</i>");
unItalicText = HtmlUtil.RemoveHtmlTags(unItalicText);
unItalicText = unItalicText.Remove(unItalicText.Length - 3, 2).TrimEnd() + " ♪";
if (ita)
unItalicText = "<i>" + unItalicText + "</i>";
}
}
if (unItalicText.StartsWith('[') && !line.StartsWith('['))
{
unItalicText = unItalicText.Remove(0, 1);
if (unItalicText.EndsWith(']'))
unItalicText = unItalicText.TrimEnd(']');
}
if (unItalicText.StartsWith('{') && !line.StartsWith('{'))
{
unItalicText = unItalicText.Remove(0, 1);
if (unItalicText.EndsWith('}'))
unItalicText = unItalicText.TrimEnd('}');
}
if (unItalicText.EndsWith('}') && !line.EndsWith('}'))
unItalicText = unItalicText.TrimEnd('}');
if (line.EndsWith("...") && unItalicText.EndsWith("”!"))
unItalicText = unItalicText.TrimEnd('!').TrimEnd('”') + ".";
if (line.EndsWith("...") && unItalicText.EndsWith("\"!"))
unItalicText = unItalicText.TrimEnd('!').TrimEnd('"') + ".";
if (line.EndsWith('.') && !unItalicText.EndsWith('.') && !unItalicText.EndsWith(".</i>"))
{
string post = string.Empty;
if (unItalicText.EndsWith("</i>"))
{
post = "</i>";
unItalicText = unItalicText.Remove(unItalicText.Length - 4);
}
if (unItalicText.EndsWith('\'') && !line.EndsWith("'."))
unItalicText = unItalicText.TrimEnd('\'');
unItalicText += "." + post;
}
if (unItalicText.EndsWith('.') && !unItalicText.EndsWith("...") && !unItalicText.EndsWith("...</i>") && line.EndsWith("..."))
{
string post = string.Empty;
if (unItalicText.EndsWith("</i>"))
{
post = "</i>";
unItalicText = unItalicText.Remove(unItalicText.Length - 4);
}
unItalicText += ".." + post;
}
if (unItalicText.EndsWith("..") && !unItalicText.EndsWith("...") && !unItalicText.EndsWith("...</i>") && line.EndsWith("..."))
{
string post = string.Empty;
if (unItalicText.EndsWith("</i>"))
{
post = "</i>";
unItalicText = unItalicText.Remove(unItalicText.Length - 4);
}
unItalicText += "." + post;
}
if (line.EndsWith('!') && !unItalicText.EndsWith('!') && !unItalicText.EndsWith("!</i>"))
{
if (unItalicText.EndsWith("!'"))
{
unItalicText = unItalicText.TrimEnd('\'');
}
else
{
if (unItalicText.EndsWith("l</i>") && _ocrFixEngine != null)
{
string w = unItalicText.Substring(0, unItalicText.Length - 4);
int wIdx = w.Length - 1;
while (wIdx >= 0 && !@" .,!?<>:;'-$@£()[]<>/""".Contains(w[wIdx]))
{
wIdx--;
}
if (wIdx + 1 < w.Length && unItalicText.Length > 5)
{
w = w.Substring(wIdx + 1);
if (!_ocrFixEngine.DoSpell(w))
unItalicText = unItalicText.Remove(unItalicText.Length - 5, 1);
}
unItalicText = unItalicText.Insert(unItalicText.Length - 4, "!");
}
else if (unItalicText.EndsWith('l') && _ocrFixEngine != null)
{
string w = unItalicText;
int wIdx = w.Length - 1;
while (wIdx >= 0 && !@" .,!?<>:;'-$@£()[]<>/""".Contains(w[wIdx]))
{
wIdx--;
}
if (wIdx + 1 < w.Length && unItalicText.Length > 5)
{
w = w.Substring(wIdx + 1);
if (!_ocrFixEngine.DoSpell(w))
unItalicText = unItalicText.Remove(unItalicText.Length - 1, 1);
}
unItalicText += "!";
}
else
{
unItalicText += "!";
}
}
}
if (line.EndsWith('?') && !unItalicText.EndsWith('?') && !unItalicText.EndsWith("?</i>"))
{
if (unItalicText.EndsWith("?'"))
unItalicText = unItalicText.TrimEnd('\'');
else
unItalicText += "?";
}
line = HtmlUtil.RemoveOpenCloseTags(unItalicText, HtmlUtil.TagItalic);
if (checkBoxAutoFixCommonErrors.Checked)
{
if (line.Contains("'.") && !textWithOutFixes.Contains("'.") && textWithOutFixes.Contains(':') && !line.EndsWith("'.") && Configuration.Settings.Tools.OcrFixUseHardcodedRules)
{
line = line.Replace("'.", ":");
}
line = _ocrFixEngine.FixOcrErrors(line, index, _lastLine, true, GetAutoGuessLevel());
}
line = "<i>" + line + "</i>";
}
else
{
unItalicText = unItalicText.Replace("</i>", string.Empty);
if (line.EndsWith("</i>", StringComparison.Ordinal) && unItalicText.EndsWith('.'))
{
line = line.Remove(line.Length - 4, 4);
if (line.EndsWith('-'))
line = line.TrimEnd('-') + ".";
if (char.IsLetter(line[line.Length - 1]))
line += ".";
line += "</i>";
}
}
}
}
}
if (checkBoxTesseractMusicOn.Checked)
{
if (line == "[J'J'J~]" || line == "[J'J'J']")
line = "[ ♪ ♪ ♪ ]";
line = line.Replace(" J' ", " ♪ ");
if (line.StartsWith("J'"))
{
line = "♪ " + line.Remove(0, 2).TrimStart();
}
if (line.StartsWith("<i>J'"))
{
line = "<i>♪ " + line.Remove(0, 5).TrimStart();
}
if (line.StartsWith("[J'"))
{
line = "[♪ " + line.Remove(0, 3).TrimStart();
}
if (line.StartsWith("<i>[J'"))
{
line = "<i>[♪ " + line.Remove(0, 6).TrimStart();
}
if (line.EndsWith("J'"))
{
line = line.Remove(line.Length - 2, 2).TrimEnd() + " ♪";
}
if (line.EndsWith("J'</i>"))
{
line = line.Remove(line.Length - 6, 6).TrimEnd() + " ♪</i>";
}
if (line.Contains(Environment.NewLine + "J'"))
{
line = line.Replace(Environment.NewLine + "J'", Environment.NewLine + "♪ ");
line = line.Replace(" ", " ");
}
if (line.Contains("J'" + Environment.NewLine))
{
line = line.Replace("J'" + Environment.NewLine, " ♪" + Environment.NewLine);
line = line.Replace(" ", " ");
}
}
if (wordsNotFound > 0 || correctWords == 0 || textWithOutFixes != null && textWithOutFixes.Replace("~", string.Empty).Trim().Length < 2)
{
_ocrFixEngine.AutoGuessesUsed.Clear();
_ocrFixEngine.UnknownWordsFound.Clear();
if (_modiEnabled && checkBoxUseModiInTesseractForUnknownWords.Checked)
{
// which is best - modi or Tesseract - we find out here
string modiText = CallModi(index);
if (modiText.Length == 0)
modiText = CallModi(index); // retry... strange MODI
if (modiText.Length == 0)
modiText = CallModi(index); // retry... strange MODI
if (modiText.Length > 1 &&
!modiText.Contains("CD") &&
(!modiText.Contains('0') || line.Contains('0')) &&
(!modiText.Contains('2') || line.Contains('2')) &&
(!modiText.Contains('3') || line.Contains('4')) &&
(!modiText.Contains('5') || line.Contains('5')) &&
(!modiText.Contains('9') || line.Contains('9')) &&
(!modiText.Contains('•') || line.Contains('•')) &&
(!modiText.Contains(')') || line.Contains(')')) &&
Utilities.CountTagInText(modiText, '(') < 2 && Utilities.CountTagInText(modiText, ')') < 2 &&
Utilities.GetNumberOfLines(modiText) < 4)
{
int modiWordsNotFound = _ocrFixEngine.CountUnknownWordsViaDictionary(modiText, out correctWords);
//if (modiWordsNotFound > 0)
{
string modiTextOcrFixed = modiText;
if (checkBoxAutoFixCommonErrors.Checked)
modiTextOcrFixed = _ocrFixEngine.FixOcrErrors(modiText, index, _lastLine, false, GetAutoGuessLevel());
int modiOcrCorrectedWordsNotFound = _ocrFixEngine.CountUnknownWordsViaDictionary(modiTextOcrFixed, out correctWords);
if (modiOcrCorrectedWordsNotFound <= modiWordsNotFound)
modiText = modiTextOcrFixed;
}
if (modiWordsNotFound < wordsNotFound || (textWithOutFixes.Length == 1 && modiWordsNotFound == 0))
line = modiText; // use the modi OCR'ed text
else if (wordsNotFound == modiWordsNotFound && modiText.EndsWith('!') && (line.EndsWith('l') || line.EndsWith('fl')))
line = modiText;
}
// take the best option - before OCR fixing, which we do again to save suggestions and prompt for user input
line = _ocrFixEngine.FixUnknownWordsViaGuessOrPrompt(out wordsNotFound, line, index, bitmap, checkBoxAutoFixCommonErrors.Checked, checkBoxPromptForUnknownWords.Checked, true, GetAutoGuessLevel());
}
else
{ // fix some error manually (modi not available)
line = _ocrFixEngine.FixUnknownWordsViaGuessOrPrompt(out wordsNotFound, line, index, bitmap, checkBoxAutoFixCommonErrors.Checked, checkBoxPromptForUnknownWords.Checked, true, GetAutoGuessLevel());
}
}
if (_ocrFixEngine.Abort)
{
ButtonStopClick(null, null);
_ocrFixEngine.Abort = false;
return string.Empty;
}
//check Tesseract... find an other way to do this...
//string tmp = HtmlUtil.RemoveHtmlTags(line).Trim();
//if (!tmp.TrimEnd().EndsWith("..."))
//{
// tmp = tmp.TrimEnd('.').TrimEnd();
// if (tmp.Length > 2 && Utilities.LowercaseLetters.Contains(tmp[tmp.Length - 1]))
// {
// if (_nocrChars == null)
// _nocrChars = LoadNOcrForTesseract("Nikse.SubtitleEdit.Resources.nOCR_TesseractHelper.xml.zip");
// string text = HtmlUtil.RemoveHtmlTags(NocrFastCheck(bitmap).TrimEnd());
// string post = string.Empty;
// if (line.EndsWith("</i>"))
// {
// post = "</i>";
// line = line.Remove(line.Length - 4, 4).Trim();
// }
// if (text.EndsWith('.'))
// {
// line = line.TrimEnd('.').Trim();
// while (text.EndsWith('.') || text.EndsWith(' '))
// {
// line += text.Substring(text.Length - 1).Trim();
// text = text.Remove(text.Length - 1, 1);
// }
// }
// else if (text.EndsWith('l') && text.EndsWith('!') && !text.EndsWith("l!"))
// {
// line = line.Remove(line.Length - 1, 1) + "!";
// }
// line += post;
// }
//}
// Log used word guesses (via word replace list)
foreach (string guess in _ocrFixEngine.AutoGuessesUsed)
listBoxLogSuggestions.Items.Add(guess);
_ocrFixEngine.AutoGuessesUsed.Clear();
// Log unkown words guess (found via spelling dictionaries)
LogUnknownWords();
ColorLineByNumberOfUnknownWords(index, wordsNotFound, line);
}
else
{ // no dictionary :(
if (checkBoxAutoFixCommonErrors.Checked)
line = _ocrFixEngine.FixOcrErrors(line, index, _lastLine, true, GetAutoGuessLevel());
if (badWords >= numberOfWords)
subtitleListView1.SetBackgroundColor(index, Color.Red);
else if (badWords >= numberOfWords / 2)
subtitleListView1.SetBackgroundColor(index, Color.Orange);
else if (badWords > 0 || line.Contains('_') || HasSingleLetters(line))
subtitleListView1.SetBackgroundColor(index, Color.Yellow);
else if (string.IsNullOrWhiteSpace(HtmlUtil.RemoveOpenCloseTags(line, HtmlUtil.TagItalic)))
subtitleListView1.SetBackgroundColor(index, Color.Orange);
else
subtitleListView1.SetBackgroundColor(index, Color.LightGreen);
}
if (textWithOutFixes.Trim() != line.Trim())
{
_tesseractOcrAutoFixes++;
labelFixesMade.Text = string.Format(" - {0}", _tesseractOcrAutoFixes);
LogOcrFix(index, textWithOutFixes, line);
}
if (_vobSubMergedPackist != null)
bitmap.Dispose();
return line;
}
