finish char bounding box

include/ocr_extractor.h

@@ -2,6 +2,7 @@
 #define OCR_EXTRACTOR_H
 
 #include <vector>
+#include <algorithm>
 #include <opencv2/opencv.hpp>
 
 using namespace std;
@@ -10,14 +11,11 @@ class OcrExtractor {
 
 public:
 
-    int ref_imgs_n {0};
-    cv::Size ref_img_size;
-    int symbol_height {0};
-    int blank_height {0};
+    int roi_idx;
+    vector<cv::Mat> roi_arr;
 
     OcrExtractor();
-    void add_ref_img(const cv::Mat & piece_img);
-
+    void add_img(const cv::Mat & piece);
 
 };
 

src/ocr_extractor.cpp

@@ -1,38 +1,28 @@
 #include <ocr_extractor.h>
 
 OcrExtractor::OcrExtractor() {
-
-}
-
-void OcrExtractor::add_ref_img(const cv::Mat & piece_img) {
 
-    if (ref_imgs_n == 0) {
-        ref_img_size = piece_img.size();
-    }
+}
 
-    ref_imgs_n++;
+void OcrExtractor::add_img(const cv::Mat & piece) {
 
     cv::Mat gray_img, col_mean;
-    cv::cvtColor(piece_img, gray_img, cv::COLOR_BGR2GRAY);
+    cv::cvtColor(piece, gray_img, cv::COLOR_BGR2GRAY);
     cv::reduce(gray_img, col_mean, 1, cv::REDUCE_AVG, CV_32FC1);
     cv::normalize(col_mean, col_mean, 0, 255, cv::NORM_MINMAX, CV_8UC1);
-    cv::threshold(col_mean, col_mean, 128, 255, cv::THRESH_BINARY);
+    cv::threshold(col_mean, col_mean, 200, 255, cv::THRESH_BINARY);
 
     int state = 0;
-    int index_st = 0;
-    int tmp_blank_height = 0;
-    int tmp_blank_n = 0;
-    int tmp_symbol_height = 0;
-    int tmp_symbol_n = 0;
+    vector<int> blank_y_arr;
+    vector<int> symbol_y_arr;
 
-    for (int y = 1; y < ref_img_size.height - 1; y++) {
+    for (int y = 0; y < piece.rows; y++) {
 
         bool stable_flag = true;
         int self_val = col_mean.at<uchar>(y, 0);
-        // cout << "self_val " << self_val << endl;
 
         for (int i = -2; i < 3; i++) {
-            if (y + i < 0 || y + i >= ref_img_size.height) continue;
+            if (y + i < 0 || y + i >= piece.rows) continue;
             if (col_mean.at<uchar>(y + i, 0) != self_val) {
                 stable_flag = false;
                 break;
@@ -42,33 +32,72 @@ void OcrExtractor::add_ref_img(const cv::Mat & piece_img) {
         if (!stable_flag) continue;
 
         switch (state) {
-            case 0:
-                index_st = y;
-                if (self_val == 0) {
+            case 0: // unknown
+                if (self_val == 255) {
                     state = -1;
+                    symbol_y_arr.push_back(y);
                 } else {
                     state = 1;
                 }
                 break;
-            case -1:
-                if (self_val == 0) continue;
-                tmp_blank_height += y - index_st;
-                tmp_blank_n++;
+            case -1: // blank
+                if (self_val == 255) continue;
+                blank_y_arr.push_back(y);
                 state = 1;
-                index_st = y;
                 break;
-            case 1:
-                if (self_val == 255) continue;
-                tmp_symbol_height += y - index_st;
-                tmp_symbol_n++;
+            case 1: // symbol
+                if (self_val == 0) continue;
+                symbol_y_arr.push_back(y);
                 state = -1;
-                index_st = y;
                 break;
         }
-        // cout << y << " " << col_mean.at<float>(y, 0) << endl;;
-        // vert_hist[y] += col_mean.at<float>(y, 0);
+
+    }
+
+    if (state == -1) blank_y_arr.push_back(piece.rows);
+
+#ifdef DEBUG
+    cv::Mat tmp_img = piece.clone();
+#endif
+
+    int blank_y_idx = 0;
+    vector<int> symbol_block_y_arr;
+    for (int symbol_y: symbol_y_arr) {
+        while (blank_y_idx < blank_y_arr.size() && blank_y_arr[blank_y_idx] <= symbol_y) blank_y_idx++;
+        int symbol_block_y = (symbol_y + blank_y_arr[blank_y_idx]) >> 1;
+        symbol_block_y_arr.push_back(symbol_block_y);
+        blank_y_idx++;
+
+#ifdef DEBUG
+        cout <<symbol_y << " " << blank_y_arr[blank_y_idx - 1] << " " << symbol_block_y << endl;
+        cv::line(tmp_img, cv::Point(0, symbol_block_y), cv::Point(piece.cols-1, symbol_block_y), cv::Scalar(0, 200, 0), 1);
+#endif
     }
 
-    cout << tmp_blank_height / tmp_blank_n << " " << tmp_symbol_height / tmp_symbol_n << endl;
+#ifdef DEBUG
+    cout << endl;
+    cv::imshow("piece", piece);
+    cv::imshow("tmp", tmp_img);
+    cv::waitKey();
+#endif
+
+    roi_arr.clear();
+    roi_idx = 0;
+    for (int i = 1; i < (int)symbol_y_arr.size(); i++) {
+
+        int y0 = symbol_block_y_arr[i - 1];
+        int y1 = symbol_block_y_arr[i];
+        int margin = int((y1 - y0) * 0.1);
+        y0 = max(0, y0 - margin);
+        y1 = min(piece.rows, y1 + margin);
+
+        int height = y1 - y0;
+        int width = min(height, piece.cols);
+        int x0 = (piece.cols - width) >> 1;
+
+        cv::Rect roi_rect(x0, y0, width, height);
+        roi_arr.push_back(piece(roi_rect).clone());
+
+    }
 
 }