feat(ocr): filter out text based on confidence

apps/server/src/services/ocr/processors/image_processor.ts

@@ -2,6 +2,7 @@ import Tesseract from 'tesseract.js';
 import { FileProcessor } from './file_processor.js';
 import { OCRResult, OCRProcessingOptions } from '../ocr_service.js';
 import log from '../../log.js';
+import options from '../../options.js';
 
 /**
  * Image processor for extracting text from image files using Tesseract
@@ -11,7 +12,7 @@ export class ImageProcessor extends FileProcessor {
     private isInitialized = false;
     private readonly supportedTypes = [
         'image/jpeg',
-        'image/jpg', 
+        'image/jpg',
         'image/png',
         'image/gif',
         'image/bmp',
@@ -42,12 +43,12 @@ export class ImageProcessor extends FileProcessor {
             // Set language if specified and different from current
             // Support multi-language format like 'ron+eng'
             const language = options.language || this.getDefaultOCRLanguage();
-            
+
             // Validate language format
             if (!this.isValidLanguageFormat(language)) {
                 throw new Error(`Invalid OCR language format: ${language}. Use format like 'eng' or 'ron+eng'`);
             }
-            
+
             if (language !== 'eng') {
                 // For different languages, create a new worker
                 await this.worker.terminate();
@@ -62,10 +63,13 @@ export class ImageProcessor extends FileProcessor {
             }
 
             const result = await this.worker.recognize(buffer);
-            
+
+            // Filter text based on minimum confidence threshold
+            const { filteredText, overallConfidence } = this.filterTextByConfidence(result.data, options);
+
             const ocrResult: OCRResult = {
-                text: result.data.text.trim(),
-                confidence: result.data.confidence / 100,  // Convert percentage to decimal
+                text: filteredText,
+                confidence: overallConfidence,
                 extractedAt: new Date().toISOString(),
                 language: options.language || this.getDefaultOCRLanguage(),
                 pageCount: 1
@@ -91,15 +95,15 @@ export class ImageProcessor extends FileProcessor {
 
         try {
             log.info('Initializing image OCR processor with Tesseract.js...');
-            
+
             // Configure proper paths for Node.js environment
             const tesseractDir = require.resolve('tesseract.js').replace('/src/index.js', '');
             const workerPath = require.resolve('tesseract.js/src/worker-script/node/index.js');
             const corePath = require.resolve('tesseract.js-core/tesseract-core.wasm.js');
-            
+
             log.info(`Using worker path: ${workerPath}`);
             log.info(`Using core path: ${corePath}`);
-            
+
             this.worker = await Tesseract.createWorker(this.getDefaultOCRLanguage(), 1, {
                 workerPath,
                 corePath,
@@ -143,6 +147,73 @@ export class ImageProcessor extends FileProcessor {
         }
     }
 
+    /**
+     * Filter text based on minimum confidence threshold
+     */
+    private filterTextByConfidence(data: any, options: OCRProcessingOptions): { filteredText: string; overallConfidence: number } {
+        const minConfidence = this.getMinConfidenceThreshold();
+
+        // If no minimum confidence set, return original text
+        if (minConfidence <= 0) {
+            return {
+                filteredText: data.text.trim(),
+                overallConfidence: data.confidence / 100
+            };
+        }
+
+        let filteredWords: string[] = [];
+        let validConfidences: number[] = [];
+
+        // Tesseract provides word-level data
+        if (data.words && Array.isArray(data.words)) {
+            for (const word of data.words) {
+                const wordConfidence = word.confidence / 100; // Convert to decimal
+
+                if (wordConfidence >= minConfidence) {
+                    filteredWords.push(word.text);
+                    validConfidences.push(wordConfidence);
+                }
+            }
+        } else {
+            // Fallback: if word-level data not available, use overall confidence
+            const overallConfidence = data.confidence / 100;
+            if (overallConfidence >= minConfidence) {
+                return {
+                    filteredText: data.text.trim(),
+                    overallConfidence
+                };
+            } else {
+                log.info(`Entire text filtered out due to low confidence ${overallConfidence} (below threshold ${minConfidence})`);
+                return {
+                    filteredText: '',
+                    overallConfidence
+                };
+            }
+        }
+
+        // Calculate average confidence of accepted words
+        const averageConfidence = validConfidences.length > 0
+            ? validConfidences.reduce((sum, conf) => sum + conf, 0) / validConfidences.length
+            : 0;
+
+        const filteredText = filteredWords.join(' ').trim();
+
+        log.info(`Filtered OCR text: ${filteredWords.length} words kept out of ${data.words?.length || 0} total words (min confidence: ${minConfidence})`);
+
+        return {
+            filteredText,
+            overallConfidence: averageConfidence
+        };
+    }
+
+    /**
+     * Get minimum confidence threshold from options
+     */
+    private getMinConfidenceThreshold(): number {
+        const minConfidence = options.getOption('ocrMinConfidence') ?? 0;
+        return parseFloat(minConfidence);
+    }
+
     /**
      * Validate OCR language format
      * Supports single language (eng) or multi-language (ron+eng)
@@ -151,16 +222,16 @@ export class ImageProcessor extends FileProcessor {
         if (!language || typeof language !== 'string') {
             return false;
         }
-        
+
         // Split by '+' for multi-language format
         const languages = language.split('+');
-        
+
         // Check each language code (should be 2-7 characters, alphanumeric with underscores)
         const validLanguagePattern = /^[a-zA-Z]{2,3}(_[a-zA-Z]{2,3})?$/;
-        
+
         return languages.every(lang => {
             const trimmed = lang.trim();
             return trimmed.length > 0 && validLanguagePattern.test(trimmed);
         });
     }
-}
+}

apps/server/src/services/options_init.ts

@@ -216,7 +216,7 @@ const defaultOptions: DefaultOption[] = [
     { name: "ocrEnabled", value: "false", isSynced: true },
     { name: "ocrLanguage", value: "eng", isSynced: true },
     { name: "ocrAutoProcessImages", value: "true", isSynced: true },
-    { name: "ocrMinConfidence", value: "0.2", isSynced: true },
+    { name: "ocrMinConfidence", value: "0.55", isSynced: true },
 ];
 
 /**