read_number_waterfill() with multiple filters

Author: jw098

SerialPrograms/Source/CommonTools/OCR/OCR_NumberReader.cpp

@@ -17,9 +17,9 @@
 #include "OCR_RawOCR.h"
 #include "OCR_NumberReader.h"
 
-// #include <iostream>
-// using std::cout;
-// using std::endl;
+#include <iostream>
+using std::cout;
+using std::endl;
 
 namespace PokemonAutomation{
 namespace OCR{
@@ -152,6 +152,82 @@ int read_number_waterfill(
     return number;
 }
 
+int read_number_waterfill(
+    Logger& logger, const ImageViewRGB32& image,
+    std::vector<std::pair<uint32_t, uint32_t>> filters,
+    uint32_t width_max,
+    bool text_inside_range
+){
+    using namespace Kernels::Waterfill;
+
+
+
+    for (std::pair<uint32_t, uint32_t> filter : filters){
+
+        uint32_t rgb32_min = filter.first;
+        uint32_t rgb32_max = filter.second;
+        
+        //  Direct OCR is unreliable. Instead, we will waterfill each character
+        //  to isolate them, then OCR them individually.
+
+        ImageRGB32 filtered = to_blackwhite_rgb32_range(image, rgb32_min, rgb32_max, text_inside_range);
+
+    //    static int c = 0;
+    //    static int i = 0;
+    //    filtered.save("test-" + std::to_string(c++) + ".png");
+
+        PackedBinaryMatrix matrix = compress_rgb32_to_binary_range(filtered, 0xff000000, 0xff7f7f7f);
+
+        std::map<size_t, WaterfillObject> map;
+        {
+            std::unique_ptr<WaterfillSession> session = make_WaterfillSession(matrix);
+            auto iter = session->make_iterator(20);
+            WaterfillObject object;
+            bool exceed_width_max = false;
+            while (map.size() < 16 && iter->find_next(object, true)){
+                if (object.width() > width_max){
+                    exceed_width_max = true;
+                    break;
+                }
+                map.emplace(object.min_x, std::move(object));
+            }
+            if (exceed_width_max){
+                // try the next color filter
+                continue;
+            }
+        }
+
+        std::string ocr_text;
+        for (const auto& item : map){
+            const WaterfillObject& object = item.second;
+            ImageRGB32 cropped = extract_box_reference(filtered, object).copy();            
+            PackedBinaryMatrix tmp(object.packed_matrix());
+            filter_by_mask(tmp, cropped, Color(0xffffffff), true);
+            ImageRGB32 padded = pad_image(cropped, cropped.width(), 0xffffffff);
+            std::string ocr = OCR::ocr_read(Language::English, padded);
+            // padded.save("test-cropped" + std::to_string(c) + "-" + std::to_string(i++) + ".png");
+            // std::cout << ocr << std::endl;
+            if (!ocr.empty()){
+                ocr_text += ocr[0];
+            }
+        }
+
+        std::string normalized = run_number_normalization(ocr_text);
+
+        if (normalized.empty()){
+            logger.log("OCR Text: \"" + ocr_text + "\" -> \"" + normalized + "\" -> Unable to read.", COLOR_RED);
+            return -1;
+        }
+
+        int number = std::atoi(normalized.c_str());
+        logger.log("OCR Text: \"" + ocr_text + "\" -> \"" + normalized + "\" -> " + std::to_string(number));
+
+        return number;
+    }
+
+    return -1;
+}
+
 
 
 

SerialPrograms/Source/CommonTools/OCR/OCR_NumberReader.h

@@ -8,6 +8,7 @@
 #define PokemonAutomation_CommonTools_OCR_NumberReader_H
 
 #include <stdint.h>
+#include <vector>
 #include "CommonFramework/Language.h"
 
 namespace PokemonAutomation{
@@ -32,6 +33,15 @@ int read_number_waterfill(
     bool text_inside_range = true
  );
 
+// applies color filters on the text, until each individual waterfilled character/number is less than width_max. then apply OCR.
+// this solves the problem where characters are too close to each other and touch, causing more than one character to be OCR'ed
+// the filters should be arranged in order of preference. (probably should be arranged broadest to narrowest)
+int read_number_waterfill(
+    Logger& logger, const ImageViewRGB32& image,
+    std::vector<std::pair<uint32_t, uint32_t>> filters,
+    uint32_t width_max,
+    bool text_inside_range = true
+ );
 
 
 }