ocr-file 0.0.2 → 0.0.6

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: f188bc0b29f4232b379e5e15d924c57a64a1758f04d8e168d2a44a744d20d1af
-  data.tar.gz: 5b54d844f01a5a5249572dd0abc270ae1fb37ff0070df9ad47eb84cf5f233fe7
+  metadata.gz: 9660e3d19c210789a7aeab56e63b002b507c12917edd1418202900d05647773b
+  data.tar.gz: 28861544f58374db141e5e3cc3ee8569201c4ffd611458e77774c98835e2f882
 SHA512:
-  metadata.gz: c51ab724a77e8b22568dc0c7cefcf3ba28407f7050976d6900824954221d4f04e677b31b58ae644c87752e60024e1667194eda8b00c89dfab30f9a81d53ba1d5
-  data.tar.gz: 9b521be6e75808899398e77cf0c0b9dee842350a5c81c0ba513ad56125725607906c8c19e6b493201750ba331521db4ba247723a1c09d82dfb61e8caec857428
+  metadata.gz: 55f4266d6877a7f2c8a4175f5601930c9940aa1a8a06fc1d6d84faca455232173f2ae4be887dc9d20246cb224498a213b91871b66ee3fb3b02699399be33453a
+  data.tar.gz: e78263baffd1ae1ff1246f705250d788ae6b5c24b3d59d78e169b3ddf93e139ac9b9688cf9d954053a913a1c3bd6be54c40466d372f915c4430d55d93282f1bb

data/Gemfile.lock

@@ -1,12 +1,13 @@
 PATH
   remote: .
   specs:
-    ocr-file (0.0.2)
+    ocr-file (0.0.6)
       active_attr (~> 0.15.4)
       console-style (~> 0.0.1)
       hexapdf (~> 0.23.0)
       mini_magick (~> 4.11.0)
       rtesseract (~> 3.1.2)
+      ruby-spellchecker (~> 0.1.5)
 
 GEM
   remote: https://rubygems.org/
@@ -60,7 +61,7 @@ GEM
       coderay (~> 1.1)
       method_source (~> 1.0)
     racc (1.6.0)
-    rack (2.2.3.1)
+    rack (2.2.4)
     rack-test (1.1.0)
       rack (>= 1.0, < 3)
     rails-dom-testing (2.0.3)
@@ -69,6 +70,7 @@ GEM
     rails-html-sanitizer (1.4.3)
       loofah (~> 2.3)
     rtesseract (3.1.2)
+    ruby-spellchecker (0.1.5)
     tzinfo (2.0.4)
       concurrent-ruby (~> 1.0)
 

data/README.md

@@ -43,13 +43,16 @@ You will need to install `tesseract` with your desired language on your system,
     ocr_engine: 'tesseract', # 'cloud-vision'
     # Image Pre-Processing
     image_preprocess: true,
-    effects: ['despeckle', 'deskew', 'enhance', 'sharpen', 'bw'], # Applies effects as listed. 'norm' is also available
+    effects: ['despeckle', 'deskew', 'enhance', 'sharpen', 'remove_shadow', 'bw'], # Applies effects as listed. 'norm' is also available
+    automatic_reprocess: true, # Will possibly do double + the operations but can produce better results automatically
     # PDF to Image Processing
     optimise_pdf: true,
     extract_pdf_images: true, # if false will screenshot each PDF page
     temp_filename_prefix: 'image',
+    spelling_correction: true, # Will attempt to fix text at the end (not used for searchable pdf output)
     # Console Output
     verbose: true,
+    timing: true,
   }
 
   doc = OcrFile::Document.new(
@@ -74,6 +77,7 @@ You will need to install `tesseract` with your desired language on your system,
   doc.to_pdf
 
   # How to merge files into a single PDF:
+  # The files can be images or other PDFs
   filepaths = []
   documents = file_paths.map { |path| OcrFile::ImageEngines::PdfEngine.open_pdf(path, password: '') }
   merged_document = OcrFile::ImageEngines::PdfEngine.merge(documents)
@@ -83,7 +87,9 @@ You will need to install `tesseract` with your desired language on your system,
 ### Notes / Tips
 Set `extract_pdf_images` to `false` for higher quality OCR. However this will consume more temporary space per PDF page and also be considerably slower.
 
-Image pre-processing only thresholds (bw), normalises the colour space, removes speckles and tries to straighten the image. Will make the end result Black and White but have far more accurate OCR (PDFs). The order of operations is important, but steps can be removed when necessary.
+Image pre-processing only thresholds (bw), normalises the colour space, removes speckles, removes shadows and tries to straighten the image. Will make the end result Black and White but have far more accurate OCR (PDFs). The order of operations is important, but steps can be removed when necessary. Expanding the colour dynamic range with `'norm'` can also be done but isn't recommended.
+
+`automatic_reprocess` is much slower as it has to re-do operations per image (in some cases) but will select the best result for each page.
 
 ### Simple CLI
 Once installed you can use `ocr-file` as a CLI. Its currently a reduced set of options. These are subject to change in future versions
@@ -108,7 +114,6 @@ To install this gem onto your local machine, run `bundle exec rake install`. To
 ### TODOs
 - input validation
 - Better CLI
-- image processing
 - password
 - Base64 encoding
 - requirements checking (installed dependencies etc ...)
@@ -117,7 +122,11 @@ To install this gem onto your local machine, run `bundle exec rake install`. To
 - Improve console output
 - Fix spaces in file names
 - Better verbosity
-- Timing
+- Docker
+- pdftk / pdf merge for text and bookmarks etc ...
+    - https://github.com/tesseract-ocr/tesseract/issues/660
+    - tesseract -c naked_pdf=true
+-
 
 ### Tests
 To run tests execute:

data/lib/ocr-file/document.rb

@@ -5,6 +5,7 @@ module OcrFile
 
     ACCEPTED_IMAGE_TYPES = ['png', 'jpeg', 'jpg', 'tiff', 'bmp']
     PAGE_BREAK = "\n\r\n" # TODO: Make configurable
+    EFFECTS_TO_REMOVE = ['', 'norm', 'remove_shadow', 'bw']
     DEFAULT_CONFIG = {
       # Images from PDF
       filetype: 'png',
@@ -22,13 +23,16 @@ module OcrFile
       ocr_engine: 'tesseract', # 'cloud-vision'
       # Image Pre-Processing
       image_preprocess: true,
-      effects: ['despeckle', 'deskew', 'enhance', 'sharpen', 'bw'],
+      effects: ['despeckle', 'deskew', 'enhance', 'sharpen', 'remove_shadow', 'bw'],
+      automatic_reprocess: true,
       # PDF to Image Processing
       optimise_pdf: true,
       extract_pdf_images: true, # if false will screenshot each PDF page
       temp_filename_prefix: 'image',
+      spelling_correction: true,
       # Console Output
       verbose: true,
+      timing: true,
     }
 
     attr_reader :original_file_path,
@@ -36,7 +40,9 @@ module OcrFile
       :save_file_path,
       :final_save_file,
       :config,
-      :ocr_engine
+      :ocr_engine,
+      :start_time,
+      :end_time
 
     # save_file_path will also generate a tmp path for tmp files. Expected folder path
     # TODO: Add in more input validation
@@ -69,72 +75,46 @@ module OcrFile
 
     # Trigger OCR pipeline
     def to_pdf
-      if pdf?
-        create_temp_folder
-        image_paths = extract_image_paths_from_pdf(@original_file_path)
-
-        pdfs_to_merge = []
-
-        image_paths.each do |image_path|
-          pdfs_to_merge << @ocr_engine.ocr_to_pdf(process_image(image_path), options: @config)
-        end
-
-        merged_pdf = OcrFile::ImageEngines::PdfEngine.merge(pdfs_to_merge)
+      @start_time = Time.now
+      find_best_image_processing(save: false) if config[:automatic_reprocess] && !text?
 
-        OcrFile::ImageEngines::PdfEngine
-          .save_pdf(merged_pdf, "#{@final_save_file}.pdf", optimise: @config[:optimise_pdf])
+      if pdf?
+        ocr_pdf_to_searchable_pdf
       elsif text?
-        text = ::OcrFile::FileHelpers.open_text_file(@original_file_path)
-        pdf_file = OcrFile::ImageEngines::PdfEngine.pdf_from_text(text, @config)
-
-        OcrFile::ImageEngines::PdfEngine
-          .save_pdf(pdf_file, "#{@final_save_file}.pdf", optimise: @config[:optimise_pdf])
+        text_to_pdf
       else # is an image
         ocr_image_to_pdf
       end
 
       close
+
+      @end_time = Time.now
+      print_time
     end
 
     def to_text
-      if pdf?
-        create_temp_folder
-        image_paths = extract_image_paths_from_pdf(@original_file_path)
-
-        image_paths.each do |image_path|
-          text = @ocr_engine.ocr_to_text(process_image(image_path), options: @config)
-          ::OcrFile::FileHelpers.append_file("#{@final_save_file}.txt", "#{text}#{PAGE_BREAK}")
-        end
-      elsif text?
-        ::OcrFile::FileHelpers.open_text_file(@original_file_path)
-      else # is an image
-        ocr_image_to_text(save: true)
-      end
+      @start_time = Time.now
+      return ::OcrFile::FileHelpers.open_text_file(@original_file_path) if text?
 
+      find_best_image_processing(save: true)
       close
+
+      @end_time = Time.now
+      print_time
     end
 
     def to_s
-      if pdf?
-        create_temp_folder
-        image_paths = extract_image_paths_from_pdf(@original_file_path)
+      @start_time = Time.now
+      return ::OcrFile::FileHelpers.open_text_file(@original_file_path) if text?
 
-        text = ''
+      text = find_best_image_processing(save: false)
 
-        image_paths.each do |image_path|
-          text = "#{text}#{PAGE_BREAK}#{@ocr_engine.ocr_to_text(process_image(image_path), options: @config)}"
-        end
+      close
 
-        close
-        text
-      elsif text?
-        ::OcrFile::FileHelpers.open_text_file(@original_file_path)
-      else # is an image
-        text = ocr_image_to_text(save: false)
+      @end_time = Time.now
+      print_time
 
-        close
-        text
-      end
+      text
     end
 
     def close
@@ -185,14 +165,67 @@ module OcrFile
       image_processor.convert!
     end
 
+    def ocr_pdf_to_searchable_pdf
+      create_temp_folder
+      image_paths = extract_image_paths_from_pdf(@original_file_path)
+
+      pdfs_to_merge = []
+
+      image_paths.each do |image_path|
+        puts image_path
+        pdfs_to_merge << @ocr_engine.ocr_to_pdf(process_image(image_path), options: @config)
+      end
+
+      merged_pdf = OcrFile::ImageEngines::PdfEngine.merge(pdfs_to_merge)
+
+      OcrFile::ImageEngines::PdfEngine
+        .save_pdf(merged_pdf, "#{@final_save_file}.pdf", optimise: @config[:optimise_pdf])
+    end
+
+    def text_to_pdf
+      text = ::OcrFile::FileHelpers.open_text_file(@original_file_path)
+      text = OcrFile::TextEngines::ResultProcessor.new(text).correct if config[:spelling_correction]
+
+      pdf_file = OcrFile::ImageEngines::PdfEngine.pdf_from_text(text, @config)
+
+      OcrFile::ImageEngines::PdfEngine
+        .save_pdf(pdf_file, "#{@final_save_file}.pdf", optimise: @config[:optimise_pdf])
+    end
+
     def ocr_image_to_pdf
+      find_best_image_processing(save: false) if config[:automatic_reprocess]
+
       pdf_document = @ocr_engine.ocr_to_pdf(process_image(@original_file_path), options: @config)
       OcrFile::ImageEngines::PdfEngine
         .save_pdf(pdf_document, "#{@final_save_file}.pdf", optimise: @config[:optimise_pdf])
     end
 
-    def ocr_image_to_text(save: true)
+    def ocr_pdf_to_text(save:)
+      create_temp_folder
+      image_paths = extract_image_paths_from_pdf(@original_file_path)
+
+      text = ''
+
+      image_paths.each do |image_path|
+        puts image_path
+        text = @ocr_engine.ocr_to_text(process_image(image_path), options: @config) || ''
+
+        text = OcrFile::TextEngines::ResultProcessor.new(text).correct if config[:spelling_correction]
+        text = "#{text}#{PAGE_BREAK}#{text}"
+      end
+
+      if save
+        ::OcrFile::FileHelpers.append_file("#{@final_save_file}.txt", "#{text}#{PAGE_BREAK}")
+      else
+        text
+      end
+    end
+
+    def ocr_image_to_text(save:)
+      create_temp_folder
+
       text = @ocr_engine.ocr_to_text(process_image(@original_file_path), options: @config)
+      text = OcrFile::TextEngines::ResultProcessor.new(text).correct if config[:spelling_correction]
 
       if save
         ::OcrFile::FileHelpers.append_file("#{@final_save_file}.txt", text)
@@ -201,6 +234,57 @@ module OcrFile
       end
     end
 
+    def ocr_file_to_text(save:)
+      if pdf?
+        ocr_pdf_to_text(save: save)
+      else # is an image
+        ocr_image_to_text(save: save)
+      end
+    end
+
+    def find_best_image_processing(save:)
+      ocr_file_to_text(save: save) unless config[:automatic_reprocess]
+
+      text = ''
+      best_text_count = 0
+      best_effects = config[:effects]
+
+      effects_to_test = [''] + (EFFECTS_TO_REMOVE - (EFFECTS_TO_REMOVE - config[:effects]))
+      effects_to_test.each do |effect|
+        text = test_ocr_settings(effect)
+        processed_result = OcrFile::TextEngines::ResultProcessor.new(text)
+
+        if processed_result.count_of_issues < best_text_count
+          best_text_count = processed_result.count_of_issues
+          best_effects = config[:effects]
+        end
+
+        break if processed_result.valid_words?
+      end
+
+      # Fallback
+      if OcrFile::TextEngines::ResultProcessor.new(text).invalid_words?
+        config[:effects] = best_effects
+        text = ocr_file_to_text(save: false)
+      end
+
+      # Adds in extra operations which is unfortunately inefficient
+      if save
+        ocr_file_to_text(save: save)
+      else
+        text
+      end
+    end
+
+    def test_ocr_settings(effect)
+      config[:effects] = config[:effects] - [effect]
+      ocr_file_to_text(save: false)
+    end
+
+    def print_time
+      puts "Total Time: #{end_time-start_time} secs.\n\n" if config[:timing]
+    end
+
     def find_ocr_engine(engine_id)
       ocr_engine_constants
         .map { |c| ocr_module(c) }

data/lib/ocr-file/image_engines/image_magick.rb

@@ -61,6 +61,13 @@ module OcrFile
         @image.sharpen('0x4') # radiusXsigma
       end
 
+      # https://github.com/ImageMagick/ImageMagick/discussions/4145
+      def remove_shadow
+        @image.negate
+        @image.lat("20x20+10\%")
+        @image.negate
+      end
+
       def deskew
         @image.deskew('40%') # threshold recommended in the docs
       end

data/lib/ocr-file/image_engines/pdf_engine.rb

@@ -61,11 +61,38 @@ module OcrFile
         image_paths
       end
 
+      def insert_image(document, image_path)
+        canvas = document.pages.add.canvas
+        canvas.image(image_path, at: [0, 0], height: 700)
+      end
+
+      def combine(text, pdf_of_images)
+        return unless pdf_of_images.is_a?(::HexaPDF::Document)
+
+        if text.is_a?(::HexaPDF::Document)
+          pages_of_text = text.pages
+        else # Assume raw text with PAGE_BREAK
+          pages_of_text = text.split(PAGE_BREAK)
+        end
+
+        return unless pages_of_text.size == pdf_of_images.pages.size
+
+        if text.is_a?(::HexaPDF::Document) # Keep the page structure
+
+        else # Just text to embed
+
+        end
+      end
+
       def merge(documents)
         target = ::HexaPDF::Document.new
 
         documents.each do |document|
-          document.pages.each { |page| target.pages << target.import(page) }
+          if document.is_a?(::HexaPDF::Document)
+            document.pages.each { |page| target.pages << target.import(page) }
+          else # Assume an image
+            insert_image(target, document)
+          end
         end
 
         target

data/lib/ocr-file/text_engines/result_processor.rb

@@ -0,0 +1,82 @@
+module OcrFile
+  module TextEngines
+    class ResultProcessor
+      MINIMUM_WORD_LENGTH = 4
+      ACCEPTABLE_NUMBER_OF_ERRORS = 8 # Random number I pulled out of nowhere
+      ACCEPTABLE_UNIDENTIFIED_WORDS = 8 # Random number I pulled out of nowhere
+
+      # REGEX
+      ASCII_ONLY = /[^\u{0000}-\u{007f}]/
+      NOISE_CHARACTERS = /[^\w\s\/-;:]/
+      DUPLICATE_WORDS = /\b(\w+)\s+\1\b/
+      EVERYTHING_BUT_CHARACTERS = /[^\w\s]|(\d)/
+
+      attr_reader :text, :clear_text
+
+      def initialize(text)
+        @text = text
+        @clear_text = generate_clear_text || text || ''
+      end
+
+      def correct
+        Spellchecker.correct(text.gsub(NOISE_CHARACTERS, '')).gsub("\n ", "\n").strip
+      end
+
+      # This is a very naive way of determining if we should re-do OCR with
+      # shifted options
+      def valid_words?
+        word_size_average >= MINIMUM_WORD_LENGTH &&
+          spelling_error_count <= ACCEPTABLE_NUMBER_OF_ERRORS &&
+          unidentified_word_count <= ACCEPTABLE_UNIDENTIFIED_WORDS
+      end
+
+      def invalid_words?
+        !valid_words?
+      end
+
+      def word_count
+        return 0 if empty_text?
+        @_word_count ||= clear_words.size
+      end
+
+      def word_size_average
+        return 0 if empty_text?
+        @_word_size_average ||= clear_words.map(&:size).sum / word_count
+      end
+
+      # Assume English
+      def unidentified_word_count
+        clear_words.reject { |word| Spellchecker::Dictionaries::EnglishWords.include?(word) }.count
+      end
+
+      def spelling_error_count
+        Spellchecker.check(clear_text).count
+      end
+
+      def count_of_issues
+        spelling_error_count + unidentified_word_count
+      end
+
+      private
+
+      def empty_text?
+        clear_text.nil? || clear_text == ''
+      end
+
+      def clear_words
+        @clear_words ||= clear_text.gsub(EVERYTHING_BUT_CHARACTERS, '').split(' ')
+      end
+
+      def generate_clear_text
+        remove_lines
+          &.gsub(ASCII_ONLY, '')
+          &.gsub(NOISE_CHARACTERS, '')
+          &.gsub(DUPLICATE_WORDS, '')
+      end
+
+      def remove_lines
+        text&.gsub("\n", ' ')&.gsub("\r", ' ')&.gsub('  ', '')
+      end
+    end
+  end
+end

data/lib/ocr-file/version.rb

@@ -1,3 +1,3 @@
 module OcrFile
-  VERSION = "0.0.2"
+  VERSION = "0.0.6"
 end

data/lib/ocr-file.rb

@@ -2,6 +2,7 @@ require 'hexapdf'
 require 'hexapdf/cli/images'
 require 'rtesseract'
 require 'mini_magick'
+require 'ruby-spellchecker'
 
 require 'ocr-file/version'
 
@@ -10,6 +11,7 @@ require 'ocr-file/image_engines/image_magick'
 require 'ocr-file/image_engines/pdftoppm'
 require 'ocr-file/ocr_engines/tesseract'
 require 'ocr-file/ocr_engines/cloud_vision'
+require 'ocr-file/text_engines/result_processor'
 require 'ocr-file/file_helpers'
 require 'ocr-file/document'
 require 'ocr-file/cli'

data/ocr-file.gemspec

@@ -32,6 +32,7 @@ Gem::Specification.new do |spec|
   spec.add_dependency "hexapdf", "~> 0.23.0"
   spec.add_dependency "rtesseract", "~> 3.1.2"
   spec.add_dependency "mini_magick", "~> 4.11.0"
+  spec.add_dependency "ruby-spellchecker", "~> 0.1.5"
 
   # Development Dependencies
   spec.add_development_dependency "pry", "~> 0.14.1"

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: ocr-file
 version: !ruby/object:Gem::Version
-  version: 0.0.2
+  version: 0.0.6
 platform: ruby
 authors:
 - trex22
 autorequire:
 bindir: bin
 cert_chain: []
-date: 2022-06-20 00:00:00.000000000 Z
+date: 2022-07-19 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: console-style
@@ -80,6 +80,20 @@ dependencies:
     - - "~>"
       - !ruby/object:Gem::Version
         version: 4.11.0
+- !ruby/object:Gem::Dependency
+  name: ruby-spellchecker
+  requirement: !ruby/object:Gem::Requirement
+    requirements:
+    - - "~>"
+      - !ruby/object:Gem::Version
+        version: 0.1.5
+  type: :runtime
+  prerelease: false
+  version_requirements: !ruby/object:Gem::Requirement
+    requirements:
+    - - "~>"
+      - !ruby/object:Gem::Version
+        version: 0.1.5
 - !ruby/object:Gem::Dependency
   name: pry
   requirement: !ruby/object:Gem::Requirement
@@ -122,6 +136,7 @@ files:
 - lib/ocr-file/image_engines/pdftoppm.rb
 - lib/ocr-file/ocr_engines/cloud_vision.rb
 - lib/ocr-file/ocr_engines/tesseract.rb
+- lib/ocr-file/text_engines/result_processor.rb
 - lib/ocr-file/version.rb
 - ocr-file.gemspec
 homepage: https://github.com/TRex22/ocr-file