do_ocr_word_mlpdo_ocr_word_mlpDoOcrWordMlpDoOcrWordMlpdo_ocr_word_mlpT_do_ocr_word_mlp
Short description
do_ocr_word_mlpdo_ocr_word_mlpDoOcrWordMlpDoOcrWordMlpdo_ocr_word_mlpT_do_ocr_word_mlp β Classify a related group of characters with an OCR classifier.
Signature
do_ocr_word_mlp( region Character, image Image, ocr_mlp OCRHandle, string Expression, integer NumAlternatives, integer NumCorrections, out string Class, out real Confidence, out string Word, out real Score )do_ocr_word_mlp( Character, Image : : OCRHandle, Expression, NumAlternatives, NumCorrections : Class, Confidence, Word, Score )void DoOcrWordMlp( const HObject& Character, const HObject& Image, const HTuple& OCRHandle, const HTuple& Expression, const HTuple& NumAlternatives, const HTuple& NumCorrections, HTuple* Class, HTuple* Confidence, HTuple* Word, HTuple* Score )static void HOperatorSet.DoOcrWordMlp( HObject character, HObject image, HTuple OCRHandle, HTuple expression, HTuple numAlternatives, HTuple numCorrections, out HTuple classVal, out HTuple confidence, out HTuple word, out HTuple score )def do_ocr_word_mlp( character: HObject, image: HObject, ocrhandle: HHandle, expression: str, num_alternatives: int, num_corrections: int ) -> Tuple[Sequence[str], Sequence[float], str, float]
def do_ocr_word_mlp_s( character: HObject, image: HObject, ocrhandle: HHandle, expression: str, num_alternatives: int, num_corrections: int ) -> Tuple[str, float, str, float]Herror T_do_ocr_word_mlp( const Hobject Character, const Hobject Image, const Htuple OCRHandle, const Htuple Expression, const Htuple NumAlternatives, const Htuple NumCorrections, Htuple* Class, Htuple* Confidence, Htuple* Word, Htuple* Score )
HTuple HRegion::DoOcrWordMlp( const HImage& Image, const HOCRMlp& OCRHandle, const HString& Expression, Hlong NumAlternatives, Hlong NumCorrections, HTuple* Confidence, HString* Word, double* Score ) const
HString HRegion::DoOcrWordMlp( const HImage& Image, const HOCRMlp& OCRHandle, const HString& Expression, Hlong NumAlternatives, Hlong NumCorrections, double* Confidence, HString* Word, double* Score ) const
HString HRegion::DoOcrWordMlp( const HImage& Image, const HOCRMlp& OCRHandle, const char* Expression, Hlong NumAlternatives, Hlong NumCorrections, double* Confidence, HString* Word, double* Score ) const
HString HRegion::DoOcrWordMlp( const HImage& Image, const HOCRMlp& OCRHandle, const wchar_t* Expression, Hlong NumAlternatives, Hlong NumCorrections, double* Confidence, HString* Word, double* Score ) const (Windows only)
HTuple HOCRMlp::DoOcrWordMlp( const HRegion& Character, const HImage& Image, const HString& Expression, Hlong NumAlternatives, Hlong NumCorrections, HTuple* Confidence, HString* Word, double* Score ) const
HString HOCRMlp::DoOcrWordMlp( const HRegion& Character, const HImage& Image, const HString& Expression, Hlong NumAlternatives, Hlong NumCorrections, double* Confidence, HString* Word, double* Score ) const
HString HOCRMlp::DoOcrWordMlp( const HRegion& Character, const HImage& Image, const char* Expression, Hlong NumAlternatives, Hlong NumCorrections, double* Confidence, HString* Word, double* Score ) const
HString HOCRMlp::DoOcrWordMlp( const HRegion& Character, const HImage& Image, const wchar_t* Expression, Hlong NumAlternatives, Hlong NumCorrections, double* Confidence, HString* Word, double* Score ) const (Windows only)
HTuple HRegion.DoOcrWordMlp( HImage image, HOCRMlp OCRHandle, string expression, int numAlternatives, int numCorrections, out HTuple confidence, out string word, out double score )
string HRegion.DoOcrWordMlp( HImage image, HOCRMlp OCRHandle, string expression, int numAlternatives, int numCorrections, out double confidence, out string word, out double score )
HTuple HOCRMlp.DoOcrWordMlp( HRegion character, HImage image, string expression, int numAlternatives, int numCorrections, out HTuple confidence, out string word, out double score )
string HOCRMlp.DoOcrWordMlp( HRegion character, HImage image, string expression, int numAlternatives, int numCorrections, out double confidence, out string word, out double score )
Description
do_ocr_word_mlpDoOcrWordMlp works like do_ocr_multi_class_mlpDoOcrMultiClassMlp insofar
as it computes the best class for each of the characters given by the
regions Charactercharactercharacter and the gray values Imageimageimage with the
OCR classifier OCRHandleOCRHandleocrhandle, and returns the classes in ClassclassValclass
and the corresponding confidences (probabilities) of the classes
in Confidenceconfidenceconfidence.
In contrast to do_ocr_multi_class_mlpDoOcrMultiClassMlp, do_ocr_word_mlpDoOcrWordMlp
treats the group of characters as an entity which yields a Wordwordword
by concatenating the class names for each character region. This allows
to restrict the allowed classification results on a textual level by
specifying an Expressionexpressionexpression describing the expected word.
The Expressionexpressionexpression may restrict the word to belong to a predefined
lexicon created using create_lexiconCreateLexicon or import_lexiconImportLexicon,
by specifying the name of the lexicon in angular
brackets as in '<mylexicon>'"<mylexicon>". If the Expressionexpressionexpression is of
any other form, it is interpreted as a regular expression with the same
syntax as specified for tuple_regexp_matchTupleRegexpMatch. Note that you will
usually want to use an expression of the form
β^β¦$β
when using variable quantifiers like β*β, to ensure that the entire word
is used in the expression. Also note that in contrast to
tuple_regexp_matchTupleRegexpMatch, do_ocr_word_mlpDoOcrWordMlp does not
support passing extra options in an expression tuple.
If the word derived from the best class for each character does not match
the Expressionexpressionexpression, do_ocr_word_mlpDoOcrWordMlp attempts to correct it by
considering the NumAlternativesnumAlternativesnum_alternatives best classes for each character.
The alternatives used are identical to those returned by
do_ocr_single_class_mlpDoOcrSingleClassMlp for a single character. It does so by
testing all possible corrections for which the classification result
is changed for at most NumCorrectionsnumCorrectionsnum_corrections character regions.
Note that NumAlternativesnumAlternativesnum_alternatives and NumCorrectionsnumCorrectionsnum_corrections affect the
complexity of the algorithm, so that in some cases internal restrictions
are made. See the section βComplexityβ below for further information.
In case the Expressionexpressionexpression is a lexicon and the above procedure did
not yield a result, the most similar word in the lexicon is returned as long
as it requires less than NumCorrectionsnumCorrectionsnum_corrections edit operations for the
correction (see suggest_lexiconSuggestLexicon).
The resulting word is graded by a Scorescorescore between 0.0 (no correction
found) and 1.0 (original word correct). The Scorescorescore is lowered by
adding a penalty according to the number of corrected characters and another
(minor) penalty depending on how many classes with higher confidences have
been ignored in order to match the Expressionexpressionexpression:
\[\begin{eqnarray*}
\textrm{Score}=1.0-\texttt{PenaltyCorrections}-\texttt{PenaltyAlternatives}
\end{eqnarray*}\]
\[\begin{eqnarray*}
\texttt{PenaltyCorrections} & = & \alpha * \texttt{num\_corr} \\
\texttt{PenaltyAlternatives} & = & \alpha * \beta * \texttt{num\_alt}
\end{eqnarray*}\]
with num_corr being the actual number of applied corrections and
num_alt the total number of discarded alternatives.
\[\begin{eqnarray*}
\alpha & = & \frac{1}{\textrm{NumCorrections}+1} \\
\beta & = & \frac{1}{\textrm{NumCorrections}*( \textrm{NumAlternatives}-1)+2}
\end{eqnarray*}\]
Note that this is a combinatorial score which does not reflect
the original Confidenceconfidenceconfidence of the best ClassclassValclass.
A string of the number
\('\backslash 032'\) (alternatively
displayed as \('\backslash 0x1A'\)) in
ClassclassValclass signifies that the region has been classified as rejection
class.
Execution information
-
Multithreading type: reentrant (runs in parallel with non-exclusive operators).
-
Multithreading scope: global (may be called from any thread).
-
Processed without parallelization.
Parameters
Charactercharactercharacter (input_object) region(-array) β objectobjectHObjectHRegionHObjectHobject
Characters to be recognized.
Imageimageimage (input_object) singlechannelimage β object (byte / uint2)object (byte / uint2)HObject (byte / uint2)HImage (byte / uint2)HObject (byte / uint2)Hobject (byte / uint2)
Gray values of the characters.
OCRHandleOCRHandleocrhandle (input_control) ocr_mlp β (handle)(handle)HTuple (HHandle)HOCRMlp, HTuple (IntPtr)HHandleHtuple (handle)
Handle of the OCR classifier.
Expressionexpressionexpression (input_control) string β (string)(string)HTuple (HString)HTuple (string)strHtuple (char*)
Expression describing the allowed word structure.
NumAlternativesnumAlternativesnum_alternatives (input_control) integer β (integer)(integer)HTuple (Hlong)HTuple (int / long)intHtuple (Hlong)
Number of classes per character considered for the
internal word correction.
Default: 33
Suggested values: 3, 4, 53, 4, 5
Value range: 1 β€ NumAlternatives
NumCorrectionsnumCorrectionsnum_corrections (input_control) integer β (integer)(integer)HTuple (Hlong)HTuple (int / long)intHtuple (Hlong)
Maximum number of corrected characters.
Default: 22
Suggested values: 1, 2, 3, 4, 51, 2, 3, 4, 5
Value range: 0 β€ NumCorrections
ClassclassValclass (output_control) string(-array) β (string)(string)HTuple (HString)HTuple (string)Sequence[str]Htuple (char*)
Result of classifying the characters with the
MLP.
Number of elements: Class == Character
Confidenceconfidenceconfidence (output_control) real(-array) β (real)(real)HTuple (double)HTuple (double)Sequence[float]Htuple (double)
Confidence of the class of the characters.
Number of elements: Confidence == Character
Wordwordword (output_control) string β (string)(string)HTuple (HString)HTuple (string)strHtuple (char*)
Word text after classification and correction.
Scorescorescore (output_control) real β (real)(real)HTuple (double)HTuple (double)floatHtuple (double)
Measure of similarity between corrected word
and uncorrected classification results.
Complexity
The complexity of checking all possible corrections is of magnitude
\(O((n*a)^{min(c,n)})\),
where \(a\) is the number of alternatives, \(n\) is the
number of character regions, and \(c\) is the number of allowed corrections.
However, to guard against a near-infinite loop in case of large \(n\),
\(c\) is internally clipped to \(5\), \(3\), or \(1\) if
\(a*n >= 30\), \(60\), or \(90\), respectively.
Result
If the parameters are valid, the operator
do_ocr_word_mlpDoOcrWordMlp returns the value 2 (H_MSG_TRUE). If
necessary, an exception is raised.
Combinations with other operators
Combinations
Possible predecessors
trainf_ocr_class_mlpTrainfOcrClassMlp, read_ocr_class_mlpReadOcrClassMlp
Alternatives
do_ocr_multi_class_mlpDoOcrMultiClassMlp
See also
create_ocr_class_mlpCreateOcrClassMlp, classify_class_mlpClassifyClassMlp
Module
OCR/OCV