gusucode.com > vision工具箱matlab源码程序 > vision/ocr.m
function txt = ocr(I, varargin)
%OCR Recognize text using Optical Character Recognition.
% txt = OCR(I) returns an ocrText object containing the recognized text,
% the location of the text within I, and a metric indicating the
% confidence of the recognition result. Confidence values range between
% 0 and 1 and should be interpreted as probabilities.
%
% txt = OCR(I, roi) recognizes text in I within one or more rectangular
% regions defined by an M-by-4 matrix, roi. Each row of roi is a
% four-element vector, [x y width height], that specifies the upper-left
% corner and size of a rectangular region of interest in pixels. Each
% rectangle must be fully contained within I.
%
% [...] = OCR(...,Name,Value) specifies additional name-value pair
% arguments described below:
%
% 'TextLayout' Specify the layout of the text within I as a string.
% Valid string values are 'Auto', 'Block', 'Line',
% 'Word', or 'Character'.
%
% Default: 'Auto'
%
% 'Language' Specify the language to recognize as a string or a
% cell array of strings. The language can be specified
% using the name of a language such as 'English' or
% 'Japanese'.
%
% <a href="matlab:helpview(fullfile(docroot,'toolbox','vision','vision.map'),'ocrLanguage')">A list of supported languages is shown in the documentation.</a>
%
% <a href="matlab:helpview(fullfile(docroot,'toolbox','vision','vision.map'),'ocrLanguage')">Custom trained languages are also supported.</a>
%
% Default: 'English'
%
% 'CharacterSet' Specify the character set as a string of characters.
% The classification process is constrained to select
% the best matches from this smaller set of characters.
% By default, all characters in the Language are used.
%
% Default: ''
%
% Class Support
% -------------
% The input image I can be logical, uint8, int16, uint16, single, or
% double, and it must be real and nonsparse.
%
% Example 1 - Recognize text within an image
% ------------------------------------------
%
% businessCard = imread('businessCard.png');
% ocrResults = OCR(businessCard)
% recognizedText = ocrResults.Text;
% figure
% imshow(businessCard)
% text(600, 150, recognizedText, ...
% 'BackgroundColor', [1 1 1])
%
% Example 2 - Recognize text in regions of interest (ROI)
% -------------------------------------------------------
% I = imread('handicapSign.jpg');
%
% % Define one or more rectangular regions of interest within I.
% roi = [360 118 384 560];
%
% % You may also use IMRECT to select a region using a mouse:
% % figure; imshow(I); roi = round(getPosition(imrect))
%
% ocrResults = OCR(I, roi);
%
% % Insert recognized text into original image
% Iocr = insertText(I, roi(1:2), ocrResults.Text, ...
% 'AnchorPoint', 'RightTop', 'FontSize',16);
% figure
% imshow(Iocr)
%
% Example 3 - Display word bounding boxes and recognition confidences
% -------------------------------------------------------------------
% businessCard = imread('businessCard.png');
% ocrResults = OCR(businessCard)
% Iocr = insertObjectAnnotation(businessCard, 'rectangle', ...
% ocrResults.WordBoundingBoxes, ...
% ocrResults.WordConfidences);
% figure
% imshow(Iocr)
%
% Example 4 - Find and highlight text in the image
% ------------------------------------------------
%
% businessCard = imread('businessCard.png');
% ocrResults = OCR(businessCard);
% bboxes = locateText(ocrResults, 'MathWorks', 'IgnoreCase', true);
% Iocr = insertShape(businessCard, 'FilledRectangle', bboxes);
% figure
% imshow(Iocr)
%
% See also ocrTrainer, ocrText, ocrText>locateText, insertShape, insertText
% References
% ----------
% An Overview of the Tesseract OCR Engine In ICDAR '07: Proceedings of the
% Ninth International Conference on Document Analysis and Recognition
% (ICDAR 2007) Vol 2 (2007), pp. 629-633 by R. Smith
%
% Ray Smith and Daria Antonova and Dar-Shyang Lee. Adapting the Tesseract
% Open Source OCR Engine for Multilingual OCR. . 2009
%
% Ray Smith. Hybrid Page Layout Analysis via Tab-Stop Detection.
% Proceedings of the 10th international conference on document analysis and
% recognition. 2009.
%#codegen
%#ok<*EMCA>
[roi, hasROI, params] = parseInputs(I,varargin{:});
if islogical(I) && ~params.PreprocessBinaryImage
% Process binary images as-is if the PreprocessBinaryImage is false.
% This by-passes tesseract's binarization stage.
[rawtext, metadata] = tesseract(params, I, roi, hasROI);
else
Iu8 = im2uint8(I);
img = vision.internal.ocr.convertRGBToGray(Iu8);
[rawtext, metadata] = tesseract(params, img, roi, hasROI);
end
txt = ocrText.create(rawtext, metadata, params);
% -------------------------------------------------------------------------
% Invoke Tesseract
% -------------------------------------------------------------------------
function [txt, ocrMetadata] = tesseract(params, Iu8, roi, hasROI)
[isSet, prefix] = unsetTessDataPrefix();
resetParameters = hasLanguageChanged(params.Language);
if vision.internal.ocr.isCodegen()
tessOpts = codegenParseParams(params);
[txt, ocrMetadata] = vision.internal.buildable.OCRBuildable.tesseract(tessOpts, Iu8, hasROI, resetParameters);
else
tessOpts = parseParams(params);
[txt, ocrMetadata] = tesseractWrapper(tessOpts, Iu8, hasROI, roi, resetParameters);
end
coder.extrinsic('setenv')
if isSimOrMex()
if isSet
setenv('TESSDATA_PREFIX',prefix)
end
end
% -------------------------------------------------------------------------
% Return true if the input language does not match the cached language.
% -------------------------------------------------------------------------
function tf = hasLanguageChanged(language)
persistent cachedLanguage
language = convertToCacheableValue(language);
% used fixed size language string to support codegen
% with max length of 4096.
n = min(4096, numel(language));
if isempty(cachedLanguage)
cachedLanguage = zeros(1,4096,'uint8');
cachedLanguage(1:n) = cast(language(1:n), 'uint8');
end
if isequal(language(1:n), cachedLanguage(1:n))
tf = false;
else
% language has changed. update cached value.
cachedLanguage(1:n) = cast(language(1:n), 'uint8');
tf = true;
end
% -------------------------------------------------------------------------
function lang = convertToCacheableValue(lang)
% cached multiple languages as concatenated string
if isempty(coder.target) && iscell(lang)
lang = [lang{:}];
end
% -------------------------------------------------------------------------
% Parse inputs.
% -------------------------------------------------------------------------
function [roi, hasROI, params] = parseInputs(I, varargin)
sz = size(I);
vision.internal.inputValidation.validateImage(I);
if mod(nargin-1,2) == 1
hasROI = true;
roi = int32(round(varargin{1}));
checkROI(roi,sz(1:2));
else
hasROI = false;
roi = ones(0,4,'int32');
end
if vision.internal.ocr.isCodegen()
if hasROI
userInput = codegenParseInputs(varargin{2:end});
else
userInput = codegenParseInputs(varargin{:});
end
else
p = getInputParser();
parse(p, varargin{:});
userInput = p.Results;
userInput.UsingCharacterSet = isempty(regexp([p.UsingDefaults{:} ''],...
'CharacterSet','once'));
end
validTextLayout = checkTextLayout(userInput.TextLayout);
[validLanguage, isCustomLanguage] = checkLanguage(userInput.Language);
if userInput.UsingCharacterSet
checkCharacterSet(userInput.CharacterSet);
end
checkPreprocessBinaryImage(userInput.PreprocessBinaryImage);
params = setParams(userInput, validLanguage, validTextLayout, isCustomLanguage);
% -------------------------------------------------------------------------
% Parse inputs during codegen.
% -------------------------------------------------------------------------
function results = codegenParseInputs(varargin)
pvPairs = struct( ...
'TextLayout', uint32(0), ...
'Language', uint32(0),...
'CharacterSet', uint32(0),...
'PreprocessBinaryImage', uint32(0));
popt = struct( ...
'CaseSensitivity', false, ...
'StructExpand' , true, ...
'PartialMatching', true);
defaults = getParamDefaults();
optarg = eml_parse_parameter_inputs(pvPairs, popt, varargin{:});
results.TextLayout = eml_get_parameter_value(optarg.TextLayout, ...
defaults.TextLayout, varargin{:});
language = coder.internal.const(eml_get_parameter_value(optarg.Language, ...
coder.internal.const(defaults.Language), varargin{:}));
results.CharacterSet = eml_get_parameter_value(optarg.CharacterSet, ...
defaults.CharacterSet, varargin{:});
results.PreprocessBinaryImage = eml_get_parameter_value(optarg.PreprocessBinaryImage, ...
defaults.PreprocessBinaryImage, varargin{:});
% UsingCharacterSet true if the user supplied one
results.UsingCharacterSet = logical(optarg.CharacterSet);
results.Language = language;
% Warning if a non-English language or custom language is specified.
if ~(optarg.Language==uint32(0)) && ~strcmpi(language,'english')
coder.internal.compileWarning('vision:ocr:codegenASCIIOnly');
end
% -------------------------------------------------------------------------
function checkROI(roi,imageSize)
for i = 1:size(roi,1)
vision.internal.detector.checkROI(roi(i,:),imageSize);
end
% -------------------------------------------------------------------------
function isCustomLang = checkIfCustomLanguage(lang)
idx = strfind(lang, '.traineddata');
if isempty(idx)
isCustomLang = false;
else
isCustomLang = true;
end
% -------------------------------------------------------------------------
function [lang, isCustomLanguage] = checkLanguage(userLanguage)
validateattributes(userLanguage,{'char','cell'},{'vector','row'}, ...
mfilename,'Language');
coder.extrinsic('exist','filesep');
if ischar(userLanguage)
isCustomLanguage = checkIfCustomLanguage(userLanguage);
if vision.internal.ocr.isCodegen()
modifiedLang = userLanguage;
else
% fix up filesep for current platform. This converts PC style \ to
% unix style / or vice versa. This prevents failures when loading
% tesseract data files.
modifiedLang = fixFilesep(userLanguage);
end
if ~isCustomLanguage
lang = vision.internal.ocr.validateLanguage(modifiedLang, vision.internal.ocr.ocrSpkgInstalled());
else
lang = modifiedLang;
if vision.internal.ocr.isCodegen()
coder.internal.errorIf(vision.internal.codegen.exist(lang) ~= 2,...
'vision:ocr:languageDataFileNotFound', lang);
else
coder.internal.errorIf(exist(lang,'file') ~= 2,...
'vision:ocr:languageDataFileNotFound', lang);
end
end
else % cell array of languages strings
% check custom language strings. When multiple custom languages are
% specified, they must all be custom languages (i.e. 'English' and a
% custom language is not allowed).
isCustomLanguage = cellfun(@(x)checkIfCustomLanguage(x),userLanguage); %#ok<EMFH>
isMixedCustomAndNonCustomLanguages = any(isCustomLanguage) && ~all(isCustomLanguage);
if isMixedCustomAndNonCustomLanguages
error(message('vision:ocr:customAndNonCustom'));
end
if all(isCustomLanguage)
% all had at least tessdata/*.traineddata
isCustomLanguage = true;
lang = fixFilesep(userLanguage);
pathstr = cell(numel(lang),1);
for i = 1:numel(lang)
pathstr{i} = fileparts(lang{i});
end
% all language data files must be in the same tessdata folder
if numel(unique(pathstr)) > 1
error(message('vision:ocr:notUniqueLanguagePath'));
% this will fail if one is a relative path and the other is a
% full path to the same folder.
end
% All language data files must be accessible
for i = 1:numel(lang)
if ~exist(lang{i},'file')
error(message('vision:ocr:languageDataFileNotFound',lang{i}));
end
end
else
% check non-custom language strings
isSupportPackageInstalled = vision.internal.ocr.ocrSpkgInstalled();
for i = 1:numel(userLanguage)
lang{i} = vision.internal.ocr.validateLanguage(userLanguage{i}, isSupportPackageInstalled);
end
isCustomLanguage = false;
end
end
% -------------------------------------------------------------------------
function checkCharacterSet(list)
validateattributes(list, {'char'},{},mfilename,'CharacterSet'); % allow empty ''
if ~isempty(list)
% make sure it's a vector
validateattributes(list, {'char'},{'vector'},mfilename,'CharacterSet');
end
% -------------------------------------------------------------------------
function str = checkTextLayout(layout)
str = validatestring(layout,{'Auto','Block','Line','Word','Character'},...
mfilename,'TextLayout');
% -------------------------------------------------------------------------
function checkPreprocessBinaryImage(value)
validateattributes(value, {'numeric','logical'}, ...
{'nonnan', 'scalar', 'real','nonsparse'}, mfilename, 'PreprocessBinaryImage');
% -------------------------------------------------------------------------
function defaults = getParamDefaults()
defaults.TextLayout = coder.internal.const('Auto');
defaults.Language = coder.internal.const('English');
defaults.CharacterSet = coder.internal.const('');
defaults.PreprocessBinaryImage = true;
% -------------------------------------------------------------------------
function params = setParams(userInput, language, textLayout, isCustomLanguage)
params.TextLayout = textLayout;
params.Language = language;
params.CharacterSet = userInput.CharacterSet;
params.UsingCharacterSet = userInput.UsingCharacterSet;
params.isCustomLanguage = coder.internal.const(isCustomLanguage);
params.PreprocessBinaryImage = logical(userInput.PreprocessBinaryImage);
% -------------------------------------------------------------------------
% Parse tesseract parameters
% -------------------------------------------------------------------------
function tessOpts = parseParams(params)
% Specify tesseract variable names as the fields of setVariable. The
% variable values should be specified as strings.
tessOpts.setVariable.tessedit_pageseg_mode = getTextLayout(params);
if params.UsingCharacterSet
tessOpts.setVariable.tessedit_char_whitelist = params.CharacterSet;
end
% enable save_blob_choices to save individual character confidence values.
tessOpts.setVariable.save_blob_choices = 'T';
[tessdata, lang] = getLanguageInfo(params);
tessOpts.tessdata = tessdata;
tessOpts.lang = lang;
% Specify tesseract initialization variables names as the fields of
% initVariable. The variable values should be specified as strings.
tessOpts.initVariable = [];
% -------------------------------------------------------------------------
% codegen: Parse tesseract parameters
% -------------------------------------------------------------------------
function tessOpts = codegenParseParams(params)
textLayout = getTextLayout(params);
if params.UsingCharacterSet
charSet = params.CharacterSet;
else
charSet = '';
end
[tessdata,lang] = getLanguageInfo(params);
tessOpts.textLayout = textLayout;
tessOpts.characterSet = charSet;
tessOpts.tessdata = tessdata;
tessOpts.lang = lang;
% -------------------------------------------------------------------------
% Return the parameter value used by tesseract to set the page segmentation
% mode (PSM). Setting other values for the page segmentation mode is not
% recommended.
% -------------------------------------------------------------------------
function textLayout = getTextLayout(params)
switch params.TextLayout
case 'Auto'
textLayout = '3';
case 'Block'
textLayout = '6';
case 'Line'
textLayout = '7';
case 'Word'
textLayout = '8';
case 'Character'
textLayout = '10';
otherwise
textLayout = ''; % codegen requires assignments for all paths
end
% -------------------------------------------------------------------------
% Return the path to the tessdata folder and the language string.
% -------------------------------------------------------------------------
function [tessdata,lang] = getLanguageInfo(params)
coder.extrinsic('ctfroot','matlabroot','fullfile','regexpi');
if params.isCustomLanguage
% params.Language contains a validated file path to the Tesseract
% language data file. The expected format is
%
% 'path/to/tessdata/foo.traineddata'
%
% where foo is the language name.
if iscell(params.Language)
% this section of code is not supported for code generation due to
% limited cell array support.
lang = cell(size(params.Language));
for i = 1:numel(params.Language)
indexStart = strfind(params.Language{i}, ['tessdata' filesep]) + 9;
indexEnd = strfind(params.Language{i}, '.traineddata') - 1;
lang{i} = params.Language{i}(indexStart(end):indexEnd(end));
end
if numel(lang) > 1
% multiple language take the form "lang1+lang2+..."
lang = strjoin(lang,'+');
else
lang = lang{1};
end
% find the string 'tessdata/foo.traineddata' located at the end of
% language data path specified by the user.
indexStart = regexpi(params.Language{1},...
'tessdata[\/\\]+(\w+)\.traineddata$','start');
tessdata = params.Language{1}(1:indexStart-1);
if isempty(tessdata)
% tessdata located in the current directory
tessdata = ['.' filesep];
end
else
tessdata = getTessdataFromPath(params.Language);
lang = getLanguageFromPath(params.Language);
end
else % a non-custom language
lang = vision.internal.ocr.convertLanguageToAlias(params.Language);
tessdata = vision.internal.ocr.locateTessdataFolder(lang);
end
% -------------------------------------------------------------------------
% Return the location of the tessdata folder from the path to a custom
% language data file.
%--------------------------------------------------------------------------
function tessdata = getTessdataFromPath(datapath)
indexStart = strfind(datapath,'tessdata');
% codegen: use isempty to check strfind result
if isempty(indexStart)
start = 0;
else
start = indexStart(end);
end
if start-1 == 0
tessdata = './';
else
tessdata = datapath(1:start-1);
end
% -------------------------------------------------------------------------
% Return language alias from the path to a custom language data file.
%--------------------------------------------------------------------------
function lang = getLanguageFromPath(datapath)
indexStart = strfind(datapath, 'tessdata') + 9;
indexEnd = strfind(datapath, '.traineddata') - 1;
if isempty(indexStart) || isempty(indexEnd)
lang = '';
else
iStart = indexStart(end);
iEnd = indexEnd(end);
lang = datapath(iStart:iEnd);
end
% -------------------------------------------------------------------------
% Return the inputParser used for parameter parsing. The inputParser is
% created once and stored in a persistent variable to improve performance.
% -------------------------------------------------------------------------
function parser = getInputParser()
persistent p;
if isempty(p)
defaults = getParamDefaults();
p = inputParser();
addOptional(p, 'ROI', []);
addParameter(p, 'TextLayout', defaults.TextLayout);
addParameter(p, 'Language', defaults.Language);
addParameter(p, 'CharacterSet', defaults.CharacterSet);
addParameter(p, 'PreprocessBinaryImage', ...
defaults.PreprocessBinaryImage);
parser = p;
else
parser = p;
end
% -------------------------------------------------------------------------
function modifiedLang = fixFilesep(userLanguage)
% Fix filesep for current platform. This converts PC style \ to unix style
% / or vice versa. This prevents failures when loading tesseract data
% files.
if vision.internal.ocr.isCodegen()
% this function is not used in codegen, but codegen requires outputs to
% be assigned on all execution paths.
modifiedLang = userLanguage;
else
modifiedLang = regexprep(userLanguage,'[\/\\]',filesep);
end
% -------------------------------------------------------------------------
% Clear the TESSDATA_PREFIX environment variable if it is set. This enables
% ocr to use the tessdata files specified using file paths instead of
% defaulting to the location of TESSDATA_PREFIX.
% -------------------------------------------------------------------------
function [isSet, prefix] = unsetTessDataPrefix()
coder.extrinsic('setenv','getenv')
if isSimOrMex()
prefix = getenv('TESSDATA_PREFIX');
if isempty(prefix)
isSet = false;
else
setenv('TESSDATA_PREFIX','');
isSet = true;
end
else
isSet = false;
prefix = '';
end
% -------------------------------------------------------------------------
% Check whether we are in sim or mex mode
% -------------------------------------------------------------------------
function tf = isSimOrMex()
tf = isempty(coder.target) || coder.target('MEX');