{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "wGp_eGfBKTEW"
   },
   "source": [
    "# The Advanced of Image Preprocessing"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "tv3vNe-kKTEd"
   },
   "source": [
    "### 本章節內容大綱\n",
    "* [影像二值化](#影像二值化)\n",
    "* [用 cv2.findContours 找影像輪廓](#用-cv2.findContours-找影像輪廓)\n",
    "* [K-Means Clustering in OpenCV](#K-Means-Clustering-in-OpenCV)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "hYExLJCEKTEd"
   },
   "outputs": [],
   "source": [
    "# opencv 在 python 中的 module 為 cv2\n",
    "import cv2\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "rZ65aUaxN4Mg"
   },
   "outputs": [],
   "source": [
    "# upload Data\n",
    "!wget -q https://github.com/TA-aiacademy/course_3.0/releases/download/CVCNN_Data/CVCNN_part1.zip\n",
    "!unzip -q CVCNN_part1.zip"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "5lWbDqm3KTEe"
   },
   "outputs": [],
   "source": [
    "image = cv2.imread(\"aia_logo.png\")[:, :, ::-1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "2GEhWeZLKTEe"
   },
   "outputs": [],
   "source": [
    "plt.imshow(image)\n",
    "plt.axis(\"off\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "J4h25mL-KTEf"
   },
   "source": [
    "* ### 影像二值化\n",
    " -- 基本上影像都是用二進位表示的，最簡單的二進位表示方式就是將一個像素值用 0(黑) 或 1(白) 表示。<br>\n",
    " -- 而將一張灰階影像轉換成黑白值可以透過 cv2.threshold 來達成。<br>\n",
    " -- 此函數的作法是給其一個門檻值，小於門檻值的皆設為黑色，反之設為白色。<br>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "Zk1ZDzV6KTEf"
   },
   "outputs": [],
   "source": [
    "gray_img = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "KjZBFH31KTEf"
   },
   "outputs": [],
   "source": [
    "#  小於 127 設為黑\n",
    "_, thresh1 = cv2.threshold(gray_img, 127, 1, cv2.THRESH_BINARY)\n",
    "\n",
    "#  小於 200 設為黑\n",
    "_, thresh2 = cv2.threshold(gray_img, 200, 1, cv2.THRESH_BINARY)\n",
    "\n",
    "\n",
    "img_list = [gray_img, thresh1, thresh2]\n",
    "title = ['Gray Image', 'Threshold: 127', 'Threshold: 200']\n",
    "\n",
    "plt.figure(figsize=(12, 12))\n",
    "for i, each in enumerate(img_list):\n",
    "    plt.subplot(1, 3, i+1)\n",
    "    plt.imshow(each, cmap='gray')\n",
    "    plt.title(title[i], fontsize=15)\n",
    "    plt.axis(\"off\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "G_XZZ0dzKTEg"
   },
   "source": [
    "- 設定門檻值，讓灰階可以黑白分明，以下為各種不同函式的表現，可以觀察原圖與所使用的函式差異。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "ZNAmpGy7T1q2"
   },
   "source": [
    "* #### Global Thresholding"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "AiVUyRubKTEg"
   },
   "outputs": [],
   "source": [
    "image = cv2.imread('bw.jpg', cv2.IMREAD_GRAYSCALE)\n",
    "(h, w) = image.shape[:2]\n",
    "center = (w/2, h/2)\n",
    "M = cv2.getRotationMatrix2D(center, 270, 2.0)\n",
    "rotated = cv2.warpAffine(image, M, (w, h))\n",
    "img = rotated\n",
    "\n",
    "_, thresh1 = cv2.threshold(img, 127, 255, cv2.THRESH_BINARY)\n",
    "_, thresh2 = cv2.threshold(img, 127, 255, cv2.THRESH_BINARY_INV)\n",
    "_, thresh3 = cv2.threshold(img, 127, 255, cv2.THRESH_TRUNC)\n",
    "_, thresh4 = cv2.threshold(img, 127, 255, cv2.THRESH_TOZERO)\n",
    "_, thresh5 = cv2.threshold(img, 127, 255, cv2.THRESH_TOZERO_INV)\n",
    "\n",
    "titles = ['Original Image', 'BINARY', 'BINARY_INV',\n",
    "          'TRUNC', 'TOZERO', 'TOZERO_INV']\n",
    "images = [img, thresh1, thresh2, thresh3, thresh4, thresh5]\n",
    "\n",
    "plt.figure(figsize=(12, 6))\n",
    "for i in range(6):\n",
    "    plt.subplot(2, 3, i+1), plt.imshow(images[i], 'gray')\n",
    "    plt.title(titles[i])\n",
    "    plt.axis(\"off\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "dC4U6VnpKTEh"
   },
   "source": [
    "[(back...)](#The-Advanced-of-Image-Preprocessing)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "ZwYaE_V2T_1o"
   },
   "source": [
    "* #### Adaptive Thresholding\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "YF804tzqKTEh"
   },
   "outputs": [],
   "source": [
    "image = cv2.imread('bicycle.jpg', cv2.IMREAD_GRAYSCALE)\n",
    "\n",
    "# 二值化(未模糊降噪)\n",
    "ret, th1 = cv2.threshold(image, 127, 255, cv2.THRESH_BINARY)\n",
    "\n",
    "# 自適應平均二值化(未模糊降噪)\n",
    "th2 = cv2.adaptiveThreshold(image, 255, cv2.ADAPTIVE_THRESH_MEAN_C,\n",
    "                            cv2.THRESH_BINARY, 11, 2)\n",
    "\n",
    "# 自適應高斯二值化(未模糊降噪)\n",
    "th3 = cv2.adaptiveThreshold(image, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C,\n",
    "                            cv2.THRESH_BINARY, 11, 2)\n",
    "\n",
    "titles = ['Original Image', 'Global Thresholding (v = 127)',\n",
    "          'Adaptive Mean Thresholding', 'Adaptive Gaussian Thresholding']\n",
    "images = [image, th1, th2, th3]\n",
    "\n",
    "for i in range(4):\n",
    "    plt.subplot(2, 2, i+1)\n",
    "    plt.imshow(images[i], 'gray')\n",
    "    plt.title(titles[i])\n",
    "    plt.xticks([])\n",
    "    plt.yticks([])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "vWrkSQ6VKTEh"
   },
   "outputs": [],
   "source": [
    "image = cv2.imread('bicycle.jpg', cv2.IMREAD_GRAYSCALE)\n",
    "\n",
    "# 將圖片做模糊化，可以降噪\n",
    "blur_img = cv2.medianBlur(image, 5)\n",
    "\n",
    "# 二值化(有模糊降噪)\n",
    "ret, th4 = cv2.threshold(blur_img, 127, 255, cv2.THRESH_BINARY)\n",
    "\n",
    "# 算術平均法的自適應二值化(有模糊降噪)\n",
    "th5 = cv2.adaptiveThreshold(blur_img, 255, cv2.ADAPTIVE_THRESH_MEAN_C,\n",
    "                            cv2.THRESH_BINARY, 11, 2)\n",
    "\n",
    "# 高斯加權均值法自適應二值化(有模糊降噪)\n",
    "th6 = cv2.adaptiveThreshold(blur_img, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C,\n",
    "                            cv2.THRESH_BINARY, 11, 2)\n",
    "\n",
    "titles = ['Blur Image', 'Global Thresholding (v = 127)',\n",
    "          'Adaptive Mean Thresholding', 'Adaptive Gaussian Thresholding']\n",
    "images = [img, th1, th2, th3]\n",
    "\n",
    "for i in range(4):\n",
    "    plt.subplot(2, 2, i+1)\n",
    "    plt.imshow(images[i], 'gray')\n",
    "    plt.title(titles[i])\n",
    "    plt.xticks([])\n",
    "    plt.yticks([])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "sp80rF8BKTEj"
   },
   "source": [
    "* ### 用 cv2.findContours 找影像輪廓"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "4U5kgf8WKTEj"
   },
   "outputs": [],
   "source": [
    "image = cv2.imread('poker.jpg')[:, :, ::-1]\n",
    "gray_img = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)\n",
    "new_image = image.copy()\n",
    "\n",
    "# threshold image\n",
    "ret, thresh = cv2.threshold(gray_img, 127, 255, cv2.THRESH_BINARY)\n",
    "# find contours and get the external one\n",
    "contours, hier = cv2.findContours(thresh, cv2.RETR_EXTERNAL,\n",
    "                                  cv2.CHAIN_APPROX_SIMPLE)\n",
    "\n",
    "\n",
    "cv2.drawContours(new_image, contours, -1, (255, 255, 0), 3)\n",
    "\n",
    "titles = ['Original_image', 'Threshed_image', 'DrawContours']\n",
    "images = [image, thresh, new_image]\n",
    "\n",
    "\n",
    "plt.figure(figsize=(18, 9))\n",
    "for i in range(3):\n",
    "    plt.subplot(2, 3, i+1),\n",
    "    plt.imshow(images[i], 'gray')\n",
    "    plt.title(titles[i])\n",
    "    plt.axis(\"off\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "gVmdpxM0KTEj"
   },
   "outputs": [],
   "source": [
    "img = cv2.imread('poker.jpg', cv2.IMREAD_UNCHANGED)\n",
    "img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)\n",
    "for i in range(4):\n",
    "    for a in range(2):\n",
    "        var_name = f'img_{i+1}{a+1}'\n",
    "        locals()[var_name] = cv2.imread('poker.jpg',\n",
    "                                        cv2.IMREAD_UNCHANGED)\n",
    "        locals()[var_name] = cv2.cvtColor(locals()[var_name],\n",
    "                                          cv2.COLOR_BGR2RGB)\n",
    "\n",
    "img_g = cv2.GaussianBlur(img, (15, 15), 5)\n",
    "\n",
    "for a in range(2):\n",
    "    # 對影像做二值化處理\n",
    "    if a == 0:\n",
    "        ret, threshed_img = cv2.threshold(\n",
    "            cv2.cvtColor(img_11, cv2.COLOR_RGB2GRAY),\n",
    "            127, 255, cv2.THRESH_BINARY\n",
    "        )\n",
    "    if a == 1:\n",
    "        ret, threshed_img = cv2.threshold(\n",
    "            cv2.cvtColor(img_g, cv2.COLOR_RGB2GRAY),\n",
    "            127, 255, cv2.THRESH_BINARY\n",
    "        )\n",
    "\n",
    "    # 找出二值化後的邊界\n",
    "    contours, hier = cv2.findContours(threshed_img,\n",
    "                                      cv2.RETR_TREE,\n",
    "                                      cv2.CHAIN_APPROX_SIMPLE)\n",
    "\n",
    "    # 沿著邊界找到小適合設定的定界框\n",
    "    # 黃：只沿著邊界畫框的\n",
    "    # 綠：沿著邊界畫出無旋轉角度的最小矩形\n",
    "    # 藍：沿著邊界畫出有旋轉角度的最小矩形\n",
    "    # 紅：沿著邊界畫出的最小圓形\n",
    "\n",
    "    for c in contours:\n",
    "        # get the bounding rect\n",
    "        x, y, w, h = cv2.boundingRect(c)\n",
    "\n",
    "        # 第一組圖\n",
    "        # draw a green rectangle to visualize the bounding rect\n",
    "        cv2.rectangle(locals()[f'img_1{a+1}'], (x, y),\n",
    "                      (x+w, y+h), (0, 255, 0), 3)\n",
    "\n",
    "        # 取最小包含物體的區域\n",
    "        rect = cv2.minAreaRect(c)\n",
    "        box = cv2.boxPoints(rect)\n",
    "\n",
    "        # 將值從 float 轉為 int\n",
    "        box = np.int0(box)\n",
    "\n",
    "        # 第二組圖\n",
    "        # draw a red 'nghien' rectangle\n",
    "        cv2.drawContours(locals()[f'img_2{a+1}'], [box],\n",
    "                         0, (0, 0, 255), 3)\n",
    "\n",
    "        (x, y), radius = cv2.minEnclosingCircle(c)\n",
    "\n",
    "        center = (int(x), int(y))\n",
    "        radius = int(radius)\n",
    "\n",
    "        # 第三組圖\n",
    "        # and draw the circle in blue\n",
    "        cv2.circle(locals()['img_3{}'.format(a+1)],\n",
    "                   center, radius, (255, 0, 0), 3)\n",
    "        # 第四組圖\n",
    "        # and draw the contour with yellow\n",
    "        cv2.drawContours(locals()['img_4{}'.format(a+1)],\n",
    "                         contours, -1, (255, 255, 0), 3)\n",
    "\n",
    "\n",
    "titles = ['Box', 'Box_blur', 'Box_2', 'Box_2_blur',\n",
    "          'Circle', 'Circle_blur', 'Contours', 'Contours_blur']\n",
    "images = [img_11, img_12, img_21, img_22,\n",
    "          img_31, img_32, img_41, img_42]\n",
    "plt.figure(figsize=(20, 20))\n",
    "for i in range(8):\n",
    "    plt.subplot(4, 2, i+1), plt.imshow(images[i])\n",
    "    plt.title(titles[i])\n",
    "    plt.axis(\"off\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "RWHSFuHzKTEj"
   },
   "source": [
    "[(back...)](#The-Advanced-of-Image-Preprocessing)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "x9shdDghKTEk"
   },
   "source": [
    "* ## K-Means Clustering in OpenCV\n",
    "一般的影像儲存方式是三個通道 (RGB)，每個通道都是 0~255(8 bits)<br>\n",
    "這樣一個像素需要花 8*3=24 bits 儲存<br>\n",
    "若想要降低影像的儲存空間，K-Means 是一種很棒的方式<br>\n",
    "我們可以將影像中的所有顏色做分群後，用群的中心點來代表這群所有點的值<br>\n",
    "將影像的顏色分成 $2^5=32$ 個群，只需要花 5 bits 就可以代表一個像素了<br>\n",
    "相對原本的儲存方式大約節省了五倍的記憶體容量了"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "_ub6Geu5KTEk"
   },
   "outputs": [],
   "source": [
    "img = cv2.imread('flower2.jpg')[:, :, ::-1]\n",
    "Z = img.reshape((-1, 3))\n",
    "\n",
    "# convert to np.float32\n",
    "Z = np.float32(Z)\n",
    "\n",
    "# define criteria, number of clusters(K) and apply kmeans()\n",
    "criteria = (cv2.TERM_CRITERIA_EPS + cv2.TERM_CRITERIA_MAX_ITER, 10, 1.0)\n",
    "\n",
    "result = []\n",
    "title = []\n",
    "\n",
    "# [不分群, 分 2 群, 分 4 群, 分 8 群, 分 16 群, 分 32 群]\n",
    "for i in range(1, 6):\n",
    "    K = 2 ** i\n",
    "    ret, label, center = cv2.kmeans(Z, K, None,\n",
    "                                    criteria, 10,\n",
    "                                    cv2.KMEANS_RANDOM_CENTERS)\n",
    "\n",
    "    center = np.uint8(center)\n",
    "    cluster = center[label.flatten()]\n",
    "    cluster = cluster.reshape((img.shape))\n",
    "\n",
    "    title .append(str(K)+\"-clustering\")\n",
    "    result.append(cluster)\n",
    "\n",
    "\n",
    "titles = ['Original_img'] + title\n",
    "images = [img] + result\n",
    "\n",
    "plt.figure(figsize=(16, 8))\n",
    "for i in range(6):\n",
    "    plt.subplot(2, 3, i+1)\n",
    "    plt.imshow(images[i])\n",
    "    plt.title(titles[i])\n",
    "    plt.axis(\"off\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "eWxMVi_BKTEk"
   },
   "source": [
    "[(back...)](#The-Advanced-of-Image-Preprocessing)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "AtOq8JhbKTEk"
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "colab": {
   "provenance": []
  },
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}