画像処理 中間発表課題

font_analyzing.pde

import processing.net.*;
import java.util.*;

Client voice_client;
PImage target_img, edge_img;
PFont font;

void setup() {
  voice_client = new Client(this, "127.0.0.1", 10500);
  target_img = loadImage("sample.jpg");
  size(target_img.width * 2, target_img.height * 3);
  edge_img = createImage(target_img.width, target_img.height, RGB);
  font = createFont("uzura_font", 140);
  background(0);

  edge_img.loadPixels();
  for(int h = 0; h < target_img.height; h++) {
    for(int w = 0; w < target_img.width; w++) {
      color c = target_img.pixels[h * target_img.width + w];
      float mid = (red(c) + green(c) + blue(c)) / 3.0;
      edge_img.pixels[h * target_img.width + w] = target_img.pixels[h * target_img.width + w];
      target_img.pixels[h * target_img.width + w] = color(mid, mid, mid);
    }
  }
  edge_img.updatePixels();
  target_img.updatePixels();
  image(target_img, 0, 0);
  
  float[][] laplacian_matrix = {{-1.0,-1.0,-1.0},{-1.0,7.2,-1.0},{-1.0,-1.0,-1.0}};
  for(int i = 0; i < 3; i++)
    for(int j = 0; j < 3; j++) laplacian_matrix[i][j] = laplacian_matrix[i][j] / 9.0;

  for(int h = 1; h < edge_img.height - 1; h++) {
    for(int w = 1; w < edge_img.width - 1; w++) {
      float filter_r = 0.0, filter_g = 0.0, filter_b = 0.0;
      
      for(int k = -1; k < 2; k++) {
        for(int j = -1; j < 2; j++) {
          color edge_col = edge_img.pixels[(h + k) * edge_img.width + (w + j)];     
          filter_r += (float)red(edge_col) * laplacian_matrix[k + 1][j + 1];
          filter_g += (float)green(edge_col) * laplacian_matrix[k + 1][j + 1];
          filter_b += (float)blue(edge_col) * laplacian_matrix[k + 1][j + 1];
        }
      }   
      float filter_mid = (filter_r + filter_g + filter_b) / 3.0;
      edge_img.pixels[h * edge_img.width + w] = filter_mid > 90 ? color(0xff,0xff,0xff) : color(0x00,0x00,0x00);
    }
  }
  image(edge_img, edge_img.width, 0);
}
HashMap<String,Integer> analyze_font(ArrayList<String> audio_al) {
  HashMap<String, Integer> res_table = new HashMap<String,Integer>();
  for(int i = 0; i < audio_al.size(); i++) {
    PGraphics pg = createGraphics(128, 128);
    pg.beginDraw();
    pg.background(0);
    pg.textFont(font);
    pg.textSize(130.0f);
    pg.text(audio_al.get(i), 15,100);
    pg.endDraw();
    
    PImage tmp_img = pg.get();
    pg.dispose();
    tmp_img.save("data/tmp/tmp_" + i + ".jpg");
  }
  for(int i = 0; i < audio_al.size(); i++) {
    int count = 0;
    PImage load_tmp_img = loadImage("data/tmp/tmp_" + i + ".jpg");
    for(int th = 0; th < load_tmp_img.height; th++) {
      for(int tw = 0; tw < load_tmp_img.width; tw++) {
        color tmp_col = load_tmp_img.pixels[th * load_tmp_img.width + tw];
        if((red(tmp_col) + green(tmp_col) + blue(tmp_col)) > ((0xff * 3) / 2)) count += 1;    
      }
    }
    res_table.put(audio_al.get(i), new Integer((int)((count / pow(128,2)) * 100)));   
  }
  return res_table;
}
ArrayList<String> getStringFromAudioInterface() {
  ArrayList<String> al = new ArrayList<String>();
  if(voice_client.available() > 0) {
    String[] splited = split(voice_client.readString(), "WORD");
    for(int i = 0; i < splited.length; i++) {
      String[] str_list = split(splited[i], "*");
      al.add(str_list[1]);
    }
  }
  return al;
}
void draw() {
  ArrayList<String> sounds = getStringFromAudioInterface();
  HashMap<String,Integer> analyze_res = analyze_font(sounds);
  /*
  print("Debug\n");
  for(String k : analyze_res.keySet()) {
    print(k + " => " + analyze_res.get(k) + "% | ");
  }
  */
  String[][] text_map = new String[target_img.height][target_img.width];
  ArrayList<Integer> sort_tmp = new ArrayList<Integer>();
  ArrayList<ArrayList<String>> sorted = new ArrayList<ArrayList<String>>();
  
  for(String dk : analyze_res.keySet()) sort_tmp.add(analyze_res.get(dk));
  Collections.sort(sort_tmp);
  for(int i = 0; i < sort_tmp.size(); i++) {
    for(String k : analyze_res.keySet()) {
      if(analyze_res.get(k) == sort_tmp.get(i)) {
        ArrayList<String> tmp_li = new ArrayList<String>();
        tmp_li.add(String.valueOf(i));
        tmp_li.add(k);
        tmp_li.add(String.valueOf(analyze_res.get(k)));
        sorted.add(tmp_li);
      }
    }
  }
  /*
  print("Debug:\n");
  for(int i = 0; i < sorted.size(); i++) {
    print("[" + sorted.get(i).get(0) + "," + sorted.get(i).get(1) + "," + sorted.get(i).get(2) + "]");
  }
  print("\n\n");
  */
  float[] pixel_ratio = new float[analyze_res.size()];
  for(int i = 0; i < analyze_res.size(); i++) pixel_ratio[i] = 255 * (1.0 / (i + 6.0));
  
  /*
  print("Debug:\n");
  for(int i = 0; i < pixel_ratio.length; i++) print(pixel_ratio[i] + " | ");
  print("\n\n");
  */
  for(int h = 0; h < target_img.height; h++) {
    for(int w = 0; w < target_img.width; w++) {
      if(edge_img.pixels[h * edge_img.width + w] == color(0xff,0xff,0xff)) {
        text_map[h][w] = sorted.get(0).get(1);
      } else if(text_map[h][w] == null) {
        float min_v = 0;
        int min_idx = 0;
        color sample = target_img.pixels[h * edge_img.width + w];
        for(int i = 1; i < pixel_ratio.length - 1; i++)  {
          min_v = Math.min(Math.abs(red(sample) - pixel_ratio[0]),Math.abs(red(sample) - pixel_ratio[i]));
        }
        for(int i = 1; i < pixel_ratio.length - 1; i++) if(Math.abs(pixel_ratio[i] - red(sample)) == min_v) min_idx = i;
        int rnd = int(random(1,2));
        text_map[h][w] = sorted.get(min_idx + rnd).get(1);
      } 
    }
  }
  String output_txt = "";
  for(int h = 0; h < target_img.height; h++) {
    for(int w = 0; w < target_img.width; w++) {
      output_txt += "  " + text_map[h][w] + " ";  
    }
    output_txt += "\n";
  }
  textSize(2);
  text(output_txt, 0,0);
}

gistfile1.sh

cd dictation-kit-v4.3.1-osx && ./configure && make
sudo make install
./bin/julius -C main.jconf -C am-gmm.jconf -module 10500