Add a command-line argument to set visible PDF image

Fix issue tesseract-ocr#210.

This adds an optional command-line argument to set the image which
will be used when generating a PDF image.

This addresses a niche case where the user wishes to use an optimized
image for OCR but maintain the visual appearance of the original image
when generating a PDF.

Signed-off-by: David Hendricks <david.hendricks@gmail.com>

api/baseapi.cpp

@@ -122,6 +122,8 @@ TessBaseAPI::TessBaseAPI()
     page_res_(NULL),
     input_file_(NULL),
     input_image_(NULL),
+    visible_pdf_image_file_(NULL),
+    visible_pdf_image_(NULL),
     output_file_(NULL),
     datapath_(NULL),
     language_(NULL),
@@ -205,6 +207,14 @@ void TessBaseAPI::SetInputName(const char* name) {
     *input_file_ = name;
 }
 
+void TessBaseAPI::SetVisiblePdfImageFilename(const char* name) {
+  if (visible_pdf_image_file_ == NULL)
+    visible_pdf_image_file_ = new STRING(name);
+  else
+    *visible_pdf_image_file_ = name;
+}
+
+
 /** Set the name of the output files. Needed only for debugging. */
 void TessBaseAPI::SetOutputName(const char* name) {
   if (output_file_ == NULL)
@@ -948,16 +958,34 @@ void TessBaseAPI::SetInputImage(Pix *pix) {
     input_image_ = pixCopy(NULL, pix);
 }
 
+void TessBaseAPI::SetVisiblePdfImage(Pix *pix) {
+  if (visible_pdf_image_)
+    pixDestroy(&visible_pdf_image_);
+  visible_pdf_image_ = NULL;
+  if (pix)
+    visible_pdf_image_ = pixCopy(NULL, pix);
+}
+
 Pix* TessBaseAPI::GetInputImage() {
   return input_image_;
 }
 
+Pix* TessBaseAPI::GetVisiblePdfImage() {
+  return visible_pdf_image_;
+}
+
 const char * TessBaseAPI::GetInputName() {
   if (input_file_)
     return input_file_->c_str();
   return NULL;
 }
 
+const char * TessBaseAPI::GetVisiblePdfImageFilename() {
+  if (visible_pdf_image_file_)
+    return visible_pdf_image_file_->c_str();
+  return NULL;
+}
+
 const char *  TessBaseAPI::GetDatapath() {
   return tesseract_->datadir.c_str();
 }
@@ -2119,6 +2147,10 @@ void TessBaseAPI::End() {
     pixDestroy(&input_image_);
     input_image_ = NULL;
   }
+  if (visible_pdf_image_ != NULL) {
+    pixDestroy(&visible_pdf_image_);
+    visible_pdf_image_ = NULL;
+  }
   if (output_file_ != NULL) {
     delete output_file_;
     output_file_ = NULL;

api/baseapi.h

@@ -146,6 +146,10 @@ class TESS_API TessBaseAPI {
   Pix* GetInputImage();
   int GetSourceYResolution();
   const char* GetDatapath();
+  void SetVisiblePdfImageFilename(const char* name);
+  const char *GetVisiblePdfImageFilename();
+  void SetVisiblePdfImage(Pix *pix);
+  Pix* GetVisiblePdfImage();
 
   /** Set the name of the bonus output files. Needed only for debugging. */
   void SetOutputName(const char* name);
@@ -871,6 +875,8 @@ class TESS_API TessBaseAPI {
   PAGE_RES*         page_res_;        ///< The page-level data.
   STRING*           input_file_;      ///< Name used by training code.
   Pix*              input_image_;     ///< Image used for searchable PDF
+  STRING*           visible_pdf_image_file_;
+  Pix*              visible_pdf_image_; ///< Image used in output PDF
   STRING*           output_file_;     ///< Name used by debug code.
   STRING*           datapath_;        ///< Current location of tessdata.
   STRING*           language_;        ///< Last initialized language.

api/pdfrenderer.cpp

@@ -841,9 +841,19 @@ bool TessPDFRenderer::imageToPDFObj(Pix *pix,
 bool TessPDFRenderer::AddImageHandler(TessBaseAPI* api) {
   size_t n;
   char buf[kBasicBufSize];
-  Pix *pix = api->GetInputImage();
   char *filename = (char *)api->GetInputName();
   int ppi = api->GetSourceYResolution();
+  Pix *pix;
+  int destroy_pix = 0;
+
+  if (api->GetVisiblePdfImageFilename()) {
+    pix = pixRead(api->GetVisiblePdfImageFilename());
+    api->SetVisiblePdfImage(pix);
+    destroy_pix = 1;
+  } else {
+    pix = api->GetInputImage();
+  }
+
   if (!pix || ppi <= 0)
     return false;
   double width = pixGetWidth(pix) * 72.0 / ppi;
@@ -915,6 +925,8 @@ bool TessPDFRenderer::AddImageHandler(TessBaseAPI* api) {
   AppendData(pdf_object, objsize);
   AppendPDFObjectDIY(objsize);
   delete[] pdf_object;
+  if (destroy_pix)
+    pixDestroy(&pix);
   return true;
 }
 

api/tesseractmain.cpp

@@ -220,6 +220,7 @@ void ParseArgs(const int argc, char** argv,
                   const char** image,
                   const char** outputbase,
                   const char** datapath,
+                  const char **visible_pdf_image_file,
                   bool* list_langs,
                   bool* print_parameters,
                   GenericVector<STRING>* vars_vec,
@@ -277,6 +278,9 @@ void ParseArgs(const int argc, char** argv,
     } else if (strcmp(argv[i], "-c") == 0 && i + 1 < argc) {
       // handled properly after api init
       ++i;
+    } else if (strcmp(argv[i], "--visible-pdf-image") == 0 && i + 1 < argc) {
+      *visible_pdf_image_file = argv[i + 1];
+      ++i;
     } else if (*image == NULL) {
       *image = argv[i];
     } else if (*outputbase == NULL) {
@@ -363,6 +367,7 @@ int main(int argc, char **argv) {
   const char* image = NULL;
   const char* outputbase = NULL;
   const char* datapath = NULL;
+  const char *visible_pdf_image_file = NULL;
   bool list_langs = false;
   bool print_parameters = false;
   GenericVector<STRING> vars_vec, vars_values;
@@ -376,6 +381,7 @@ int main(int argc, char **argv) {
 
   ParseArgs(argc, argv,
           &lang, &image, &outputbase, &datapath,
+          &visible_pdf_image_file,
           &list_langs, &print_parameters,
           &vars_vec, &vars_values, &arg_i, &pagesegmode);
 
@@ -414,6 +420,9 @@ int main(int argc, char **argv) {
 
   FixPageSegMode(&api, pagesegmode);
 
+  if (visible_pdf_image_file)
+    api.SetVisiblePdfImageFilename(visible_pdf_image_file);
+
   if (pagesegmode == tesseract::PSM_AUTO_ONLY) {
     int ret_val = 0;