tesseract v5.3.3.20231005
Public Member Functions | Static Public Member Functions | List of all members
tesseract::ImageData Class Reference

#include <imagedata.h>

Public Member Functions

 ImageData ()
 
 ImageData (bool vertical, Image pix)
 
 ~ImageData ()
 
bool Serialize (TFile *fp) const
 
bool DeSerialize (TFile *fp)
 
const std::string & imagefilename () const
 
void set_imagefilename (const std::string &name)
 
int page_number () const
 
void set_page_number (int num)
 
const std::vector< char > & image_data () const
 
const std::string & language () const
 
void set_language (const std::string &lang)
 
const std::string & transcription () const
 
const std::vector< TBOX > & boxes () const
 
const std::vector< std::string > & box_texts () const
 
const std::string & box_text (int index) const
 
void SetPix (Image pix)
 
Image GetPix () const
 
Image PreScale (int target_height, int max_height, float *scale_factor, int *scaled_width, int *scaled_height, std::vector< TBOX > *boxes) const
 
int MemoryUsed () const
 
void Display () const
 
void AddBoxes (const std::vector< TBOX > &boxes, const std::vector< std::string > &texts, const std::vector< int > &box_pages)
 

Static Public Member Functions

static ImageData * Build (const char *name, int page_number, const char *lang, const char *imagedata, int imagedatasize, const char *truth_text, const char *box_text)
 
static bool SkipDeSerialize (TFile *fp)
 

Detailed Description

Definition at line 62 of file imagedata.h.

Constructor & Destructor Documentation

◆ ImageData() [1/2]

tesseract::ImageData::ImageData ( )

Definition at line 44 of file imagedata.cpp.

44: page_number_(-1), vertical_text_(false) {}

◆ ImageData() [2/2]

tesseract::ImageData::ImageData ( bool  vertical,
Image  pix 
)

Definition at line 46 of file imagedata.cpp.

47 : page_number_(0), vertical_text_(vertical) {
48 SetPix(pix);
49}
tesseract::ImageData::SetPix
void SetPix(Image pix)
Definition: imagedata.cpp:184

◆ ~ImageData()

tesseract::ImageData::~ImageData ( )

Definition at line 50 of file imagedata.cpp.

50 {
51#ifdef TESSERACT_IMAGEDATA_AS_PIX
52 internal_pix_.destroy();
53#endif
54}

Member Function Documentation

◆ AddBoxes()

void tesseract::ImageData::AddBoxes ( const std::vector< TBOX > &  boxes,
const std::vector< std::string > &  texts,
const std::vector< int > &  box_pages 
)

Definition at line 315 of file imagedata.cpp.

317 {
318 // Copy the boxes and make the transcription.
319 for (unsigned i = 0; i < box_pages.size(); ++i) {
320 if (page_number_ >= 0 && box_pages[i] != page_number_) {
321 continue;
322 }
323 transcription_ += texts[i];
324 boxes_.push_back(boxes[i]);
325 box_texts_.push_back(texts[i]);
326 }
327}
tesseract::ImageData::boxes
const std::vector< TBOX > & boxes() const
Definition: imagedata.h:107

◆ box_text()

const std::string & tesseract::ImageData::box_text ( int  index) const
inline

Definition at line 113 of file imagedata.h.

113 {
114 return box_texts_[index];
115 }

◆ box_texts()

const std::vector< std::string > & tesseract::ImageData::box_texts ( ) const
inline

Definition at line 110 of file imagedata.h.

110 {
111 return box_texts_;
112 }

◆ boxes()

const std::vector< TBOX > & tesseract::ImageData::boxes ( ) const
inline

Definition at line 107 of file imagedata.h.

107 {
108 return boxes_;
109 }

◆ Build()

ImageData * tesseract::ImageData::Build ( const char *  name,
int  page_number,
const char *  lang,
const char *  imagedata,
int  imagedatasize,
const char *  truth_text,
const char *  box_text 
)
static

Definition at line 58 of file imagedata.cpp.

60 {
61 auto *image_data = new ImageData();
62 image_data->imagefilename_ = name;
63 image_data->page_number_ = page_number;
64 image_data->language_ = lang;
65 // Save the imagedata.
66 // TODO: optimize resize (no init).
67 image_data->image_data_.resize(imagedatasize);
68 memcpy(&image_data->image_data_[0], imagedata, imagedatasize);
69 if (!image_data->AddBoxes(box_text)) {
70 if (truth_text == nullptr || truth_text[0] == '\0') {
71 tprintf("Error: No text corresponding to page %d from image %s!\n",
72 page_number, name);
73 delete image_data;
74 return nullptr;
75 }
76 image_data->transcription_ = truth_text;
77 // If we have no boxes, the transcription is in the 0th box_texts_.
78 image_data->box_texts_.emplace_back(truth_text);
79 // We will create a box for the whole image on PreScale, to save unpacking
80 // the image now.
81 } else if (truth_text != nullptr && truth_text[0] != '\0' &&
82 image_data->transcription_ != truth_text) {
83 // Save the truth text as it is present and disagrees with the box text.
84 image_data->transcription_ = truth_text;
85 }
86 return image_data;
87}
tesseract::tprintf
void tprintf(const char *format,...)
Definition: tprintf.cpp:41
tesseract::ImageData::page_number
int page_number() const
Definition: imagedata.h:89
tesseract::ImageData::image_data
const std::vector< char > & image_data() const
Definition: imagedata.h:95
tesseract::ImageData::box_text
const std::string & box_text(int index) const
Definition: imagedata.h:113

◆ DeSerialize()

bool tesseract::ImageData::DeSerialize ( TFile *  fp)

Definition at line 117 of file imagedata.cpp.

117 {
118 if (!fp->DeSerialize(imagefilename_)) {
119 return false;
120 }
121 if (!fp->DeSerialize(&page_number_)) {
122 return false;
123 }
124 if (!fp->DeSerialize(image_data_)) {
125 return false;
126 }
127 if (!fp->DeSerialize(language_)) {
128 return false;
129 }
130 if (!fp->DeSerialize(transcription_)) {
131 return false;
132 }
133 if (!fp->DeSerialize(boxes_)) {
134 return false;
135 }
136 if (!fp->DeSerialize(box_texts_)) {
137 return false;
138 }
139 int8_t vertical = 0;
140 if (!fp->DeSerialize(&vertical)) {
141 return false;
142 }
143 vertical_text_ = vertical != 0;
144 return true;
145}

◆ Display()

void tesseract::ImageData::Display ( ) const

Definition at line 275 of file imagedata.cpp.

275 {
276 const int kTextSize = 64;
277 // Draw the image.
278 Image pix = GetPix();
279 if (pix == nullptr) {
280 return;
281 }
282 int width = pixGetWidth(pix);
283 int height = pixGetHeight(pix);
284 auto *win = new ScrollView("Imagedata", 100, 100, 2 * (width + 2 * kTextSize),
285 2 * (height + 4 * kTextSize), width + 10,
286 height + 3 * kTextSize, true);
287 win->Draw(pix, 0, height - 1);
288 pix.destroy();
289 // Draw the boxes.
290 win->Pen(ScrollView::RED);
291 win->Brush(ScrollView::NONE);
292 int text_size = kTextSize;
293 if (!boxes_.empty() && boxes_[0].height() * 2 < text_size) {
294 text_size = boxes_[0].height() * 2;
295 }
296 win->TextAttributes("Arial", text_size, false, false, false);
297 if (!boxes_.empty()) {
298 for (unsigned b = 0; b < boxes_.size(); ++b) {
299 boxes_[b].plot(win);
300 win->Text(boxes_[b].left(), height + kTextSize, box_texts_[b].c_str());
301 }
302 } else {
303 // The full transcription.
304 win->Pen(ScrollView::CYAN);
305 win->Text(0, height + kTextSize * 2, transcription_.c_str());
306 }
307 win->Update();
308 win->Wait();
309}
tesseract::ImageData::GetPix
Image GetPix() const
Definition: imagedata.cpp:193

◆ GetPix()

Image tesseract::ImageData::GetPix ( ) const

Definition at line 193 of file imagedata.cpp.

193 {
194#ifdef TESSERACT_IMAGEDATA_AS_PIX
195# ifdef GRAPHICS_DISABLED
196 /* The only caller of this is the scaling functions to prescale the
197 * source. Thus we can just return a new pointer to the same data. */
198 return internal_pix_.clone();
199# else
200 /* pixCopy always does an actual copy, so the caller can modify the
201 * changed data. */
202 return internal_pix_.copy();
203# endif
204#else
205 return GetPixInternal(image_data_);
206#endif
207}

◆ image_data()

const std::vector< char > & tesseract::ImageData::image_data ( ) const
inline

Definition at line 95 of file imagedata.h.

95 {
96 return image_data_;
97 }

◆ imagefilename()

const std::string & tesseract::ImageData::imagefilename ( ) const
inline

Definition at line 83 of file imagedata.h.

83 {
84 return imagefilename_;
85 }

◆ language()

const std::string & tesseract::ImageData::language ( ) const
inline

Definition at line 98 of file imagedata.h.

98 {
99 return language_;
100 }

◆ MemoryUsed()

int tesseract::ImageData::MemoryUsed ( ) const

Definition at line 268 of file imagedata.cpp.

268 {
269 return image_data_.size();
270}

◆ page_number()

int tesseract::ImageData::page_number ( ) const
inline

Definition at line 89 of file imagedata.h.

89 {
90 return page_number_;
91 }

◆ PreScale()

Image tesseract::ImageData::PreScale ( int  target_height,
int  max_height,
float *  scale_factor,
int *  scaled_width,
int *  scaled_height,
std::vector< TBOX > *  boxes 
) const

Definition at line 215 of file imagedata.cpp.

217 {
218 int input_width = 0;
219 int input_height = 0;
220 Image src_pix = GetPix();
221 ASSERT_HOST(src_pix != nullptr);
222 input_width = pixGetWidth(src_pix);
223 input_height = pixGetHeight(src_pix);
224 if (target_height == 0) {
225 target_height = std::min(input_height, max_height);
226 }
227 float im_factor = static_cast<float>(target_height) / input_height;
228 if (scaled_width != nullptr) {
229 *scaled_width = IntCastRounded(im_factor * input_width);
230 }
231 if (scaled_height != nullptr) {
232 *scaled_height = target_height;
233 }
234 // Get the scaled image.
235 Image pix = pixScale(src_pix, im_factor, im_factor);
236 if (pix == nullptr) {
237 tprintf("Scaling pix of size %d, %d by factor %g made null pix!!\n",
238 input_width, input_height, im_factor);
239 src_pix.destroy();
240 return nullptr;
241 }
242 if (scaled_width != nullptr) {
243 *scaled_width = pixGetWidth(pix);
244 }
245 if (scaled_height != nullptr) {
246 *scaled_height = pixGetHeight(pix);
247 }
248 src_pix.destroy();
249 if (boxes != nullptr) {
250 // Get the boxes.
251 boxes->clear();
252 for (auto box : boxes_) {
253 box.scale(im_factor);
254 boxes->push_back(box);
255 }
256 if (boxes->empty()) {
257 // Make a single box for the whole image.
258 TBOX box(0, 0, im_factor * input_width, target_height);
259 boxes->push_back(box);
260 }
261 }
262 if (scale_factor != nullptr) {
263 *scale_factor = im_factor;
264 }
265 return pix;
266}
ASSERT_HOST
#define ASSERT_HOST(x)
Definition: errcode.h:54
TBOX
@ TBOX
Definition: cleanapi_test.cc:19
tesseract::IntCastRounded
int IntCastRounded(double x)
Definition: helpers.h:170

◆ Serialize()

bool tesseract::ImageData::Serialize ( TFile *  fp) const

Definition at line 90 of file imagedata.cpp.

90 {
91 if (!fp->Serialize(imagefilename_)) {
92 return false;
93 }
94 if (!fp->Serialize(&page_number_)) {
95 return false;
96 }
97 if (!fp->Serialize(image_data_)) {
98 return false;
99 }
100 if (!fp->Serialize(language_)) {
101 return false;
102 }
103 if (!fp->Serialize(transcription_)) {
104 return false;
105 }
106 if (!fp->Serialize(boxes_)) {
107 return false;
108 }
109 if (!fp->Serialize(box_texts_)) {
110 return false;
111 }
112 int8_t vertical = vertical_text_;
113 return fp->Serialize(&vertical);
114}

◆ set_imagefilename()

void tesseract::ImageData::set_imagefilename ( const std::string &  name)
inline

Definition at line 86 of file imagedata.h.

86 {
87 imagefilename_ = name;
88 }

◆ set_language()

void tesseract::ImageData::set_language ( const std::string &  lang)
inline

Definition at line 101 of file imagedata.h.

101 {
102 language_ = lang;
103 }

◆ set_page_number()

void tesseract::ImageData::set_page_number ( int  num)
inline

Definition at line 92 of file imagedata.h.

92 {
93 page_number_ = num;
94 }

◆ SetPix()

void tesseract::ImageData::SetPix ( Image  pix)

Definition at line 184 of file imagedata.cpp.

184 {
185#ifdef TESSERACT_IMAGEDATA_AS_PIX
186 internal_pix_ = pix;
187#else
188 SetPixInternal(pix, &image_data_);
189#endif
190}

◆ SkipDeSerialize()

bool tesseract::ImageData::SkipDeSerialize ( TFile *  fp)
static

Definition at line 148 of file imagedata.cpp.

148 {
149 if (!fp->DeSerializeSkip()) {
150 return false;
151 }
152 int32_t page_number;
153 if (!fp->DeSerialize(&page_number)) {
154 return false;
155 }
156 if (!fp->DeSerializeSkip()) {
157 return false;
158 }
159 if (!fp->DeSerializeSkip()) {
160 return false;
161 }
162 if (!fp->DeSerializeSkip()) {
163 return false;
164 }
165 if (!fp->DeSerializeSkip(sizeof(TBOX))) {
166 return false;
167 }
168 int32_t number;
169 if (!fp->DeSerialize(&number)) {
170 return false;
171 }
172 for (int i = 0; i < number; i++) {
173 if (!fp->DeSerializeSkip()) {
174 return false;
175 }
176 }
177 int8_t vertical = 0;
178 return fp->DeSerialize(&vertical);
179}

◆ transcription()

const std::string & tesseract::ImageData::transcription ( ) const
inline

Definition at line 104 of file imagedata.h.

104 {
105 return transcription_;
106 }
The documentation for this class was generated from the following files: