File size: 9,074 Bytes
d520909
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
"""
Tesseract OCR Engine

Fallback OCR engine using Tesseract.
Provides broad language support and is widely available.
"""

import time
from typing import List, Optional, Dict, Any
import numpy as np
from loguru import logger

from .base import OCREngine, OCRConfig, OCRResult
from ..schemas.core import BoundingBox, OCRRegion

# Try to import pytesseract
try:
    import pytesseract
    from PIL import Image
    HAS_TESSERACT = True
except ImportError:
    HAS_TESSERACT = False
    logger.warning(
        "pytesseract not installed. Install with: pip install pytesseract "
        "Also install Tesseract: apt-get install tesseract-ocr"
    )


class TesseractOCREngine(OCREngine):
    """
    OCR engine using Tesseract.

    Features:
    - Broad language support (100+ languages)
    - Mature and well-tested
    - No GPU required
    - Page segmentation modes for different layouts
    """

    # Tesseract language codes (subset of common ones)
    LANGUAGE_MAP = {
        "en": "eng",
        "ch": "chi_sim",
        "chinese_cht": "chi_tra",
        "fr": "fra",
        "german": "deu",
        "es": "spa",
        "it": "ita",
        "pt": "por",
        "ru": "rus",
        "japan": "jpn",
        "korean": "kor",
        "ar": "ara",
        "hi": "hin",
        "latin": "lat",
    }

    # Page segmentation modes
    PSM_AUTO = 3  # Fully automatic page segmentation
    PSM_SINGLE_BLOCK = 6  # Assume single uniform block of text
    PSM_SINGLE_LINE = 7  # Treat image as single line
    PSM_SPARSE = 11  # Sparse text with no particular order

    def __init__(self, config: Optional[OCRConfig] = None):
        """Initialize Tesseract OCR engine."""
        super().__init__(config)
        self._tesseract_cmd: Optional[str] = None

    def initialize(self):
        """Initialize Tesseract engine."""
        if not HAS_TESSERACT:
            raise RuntimeError(
                "pytesseract not installed. Install with: pip install pytesseract. "
                "Also install Tesseract: apt-get install tesseract-ocr"
            )

        if self._initialized:
            return

        logger.info("Initializing Tesseract OCR engine...")

        # Test Tesseract installation
        try:
            version = pytesseract.get_tesseract_version()
            logger.info(f"Tesseract version: {version}")
            self._initialized = True
        except Exception as e:
            logger.error(f"Tesseract not properly installed: {e}")
            raise RuntimeError(
                f"Tesseract not properly installed: {e}. "
                "Install with: apt-get install tesseract-ocr"
            )

    def recognize(
        self,
        image: np.ndarray,
        page_number: int = 0,
    ) -> OCRResult:
        """
        Perform OCR on an image using Tesseract.

        Args:
            image: Image as numpy array (RGB, HWC format)
            page_number: Page number for multi-page documents

        Returns:
            OCRResult with recognized text and regions
        """
        if not self._initialized:
            self.initialize()

        start_time = time.time()

        try:
            # Convert numpy array to PIL Image
            pil_image = Image.fromarray(image)

            # Build language string
            lang = self._get_tesseract_lang()

            # Configure Tesseract
            custom_config = self._build_config()

            # Get detailed data with bounding boxes
            data = pytesseract.image_to_data(
                pil_image,
                lang=lang,
                config=custom_config,
                output_type=pytesseract.Output.DICT,
            )

            # Process results
            regions = []
            all_texts = []
            total_confidence = 0.0
            valid_count = 0

            height, width = image.shape[:2]

            # Group words into lines
            current_line_id = -1
            word_id = 0

            for i in range(len(data['text'])):
                text = data['text'][i].strip()
                conf = int(data['conf'][i])

                # Skip empty or low confidence
                if not text or conf < 0:
                    continue

                confidence = conf / 100.0
                if confidence < self.config.min_confidence:
                    continue

                # Track line changes
                block_num = data['block_num'][i]
                line_num = data['line_num'][i]
                line_id = block_num * 1000 + line_num

                if line_id != current_line_id:
                    current_line_id = line_id
                    word_id = 0
                else:
                    word_id += 1

                # Get bounding box
                x = data['left'][i]
                y = data['top'][i]
                w = data['width'][i]
                h = data['height'][i]

                bbox = BoundingBox(
                    x_min=float(x),
                    y_min=float(y),
                    x_max=float(x + w),
                    y_max=float(y + h),
                    normalized=False,
                    page_width=width,
                    page_height=height,
                )

                region = OCRRegion(
                    text=text,
                    confidence=confidence,
                    bbox=bbox,
                    page=page_number,
                    line_id=line_id,
                    word_id=word_id,
                    engine="tesseract",
                )
                regions.append(region)
                all_texts.append(text)
                total_confidence += confidence
                valid_count += 1

            # Also get full text for better formatting
            full_text = pytesseract.image_to_string(
                pil_image,
                lang=lang,
                config=custom_config,
            )

            processing_time = (time.time() - start_time) * 1000

            return OCRResult(
                regions=regions,
                full_text=full_text.strip(),
                confidence_avg=total_confidence / valid_count if valid_count > 0 else 0.0,
                processing_time_ms=processing_time,
                engine="tesseract",
                success=True,
            )

        except Exception as e:
            logger.error(f"Tesseract recognition failed: {e}")
            return OCRResult(
                regions=[],
                full_text="",
                confidence_avg=0.0,
                processing_time_ms=(time.time() - start_time) * 1000,
                engine="tesseract",
                success=False,
                error=str(e),
            )

    def _get_tesseract_lang(self) -> str:
        """Get Tesseract language string from config."""
        langs = []
        for lang in self.config.languages:
            tess_lang = self.LANGUAGE_MAP.get(lang, "eng")
            if tess_lang not in langs:
                langs.append(tess_lang)
        return "+".join(langs) if langs else "eng"

    def _build_config(self) -> str:
        """Build Tesseract config string."""
        config_parts = [
            f"--psm {self.PSM_AUTO}",  # Page segmentation mode
            "--oem 3",  # Use both legacy and LSTM engines
        ]

        # Add more options as needed
        if self.config.return_word_boxes:
            config_parts.append("-c preserve_interword_spaces=1")

        return " ".join(config_parts)

    def get_supported_languages(self) -> List[str]:
        """Return list of supported language codes."""
        return list(self.LANGUAGE_MAP.keys())

    def get_installed_languages(self) -> List[str]:
        """Get list of languages installed in Tesseract."""
        if not self._initialized:
            self.initialize()

        try:
            langs = pytesseract.get_languages()
            return langs
        except Exception as e:
            logger.warning(f"Could not get installed languages: {e}")
            return ["eng"]

    def recognize_with_hocr(
        self,
        image: np.ndarray,
        page_number: int = 0,
    ) -> tuple:
        """
        Perform OCR and return hOCR format for detailed layout.

        Args:
            image: Image as numpy array
            page_number: Page number

        Returns:
            Tuple of (OCRResult, hOCR string)
        """
        if not self._initialized:
            self.initialize()

        pil_image = Image.fromarray(image)
        lang = self._get_tesseract_lang()
        config = self._build_config()

        # Get standard result
        ocr_result = self.recognize(image, page_number)

        # Get hOCR for layout analysis
        try:
            hocr = pytesseract.image_to_pdf_or_hocr(
                pil_image,
                lang=lang,
                config=config,
                extension='hocr',
            )
            return ocr_result, hocr.decode('utf-8')
        except Exception as e:
            logger.warning(f"Failed to generate hOCR: {e}")
            return ocr_result, None