Pre-Processing.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "CSC420_Project.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "code",
      "execution_count": 0,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 68
        },
        "colab_type": "code",
        "id": "zKGkNXBYQA-O",
        "outputId": "71e226e0-3c28-4c79-e5dc-5b3e0a681d73"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Drive already mounted at /content/drive; to attempt to forcibly remount, call drive.mount(\"/content/drive\", force_remount=True).\n",
            "Requirement already satisfied: opencv-python in /usr/local/lib/python3.6/dist-packages (3.4.7.28)\n",
            "Requirement already satisfied: numpy>=1.11.3 in /usr/local/lib/python3.6/dist-packages (from opencv-python) (1.17.4)\n"
          ]
        }
      ],
      "source": [
        "# Load the Drive helper and mount\n",
        "from google.colab import drive\n",
        "\n",
        "# This will prompt for authorization.\n",
        "drive.mount('/content/drive')\n",
        "\n",
        "# install OpenCV\n",
        "!pip install opencv-python\n",
        "\n",
        "# the usual ...\n",
        "%matplotlib inline\n",
        "import matplotlib.pyplot as plt\n",
        "import cv2\n",
        "import numpy as np\n",
        "import matplotlib.image as mpimg\n",
        "import math"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 0,
      "metadata": {
        "colab": {},
        "colab_type": "code",
        "id": "eHhZ8xGPYOHo"
      },
      "outputs": [],
      "source": [
        "########################################################################\n",
        "# Converting image to greyscale\n",
        "########################################################################\n",
        "\n",
        "def convert_to_greyscale(color_img):\n",
        "    grey_img = np.dot(\n",
        "        color_img[..., :3], \n",
        "        [0.299, 0.587, 0.144])\n",
        "    return grey_img"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 0,
      "metadata": {
        "colab": {},
        "colab_type": "code",
        "id": "v34280-VZ7x6"
      },
      "outputs": [],
      "source": [
        "########################################################################\n",
        "# Applying blur to the image\n",
        "########################################################################\n",
        "\n",
        "def apply_correlation(img, kernel, mode='valid'):\n",
        "    img = cv2.imread(img, 0)\n",
        "    img_height, img_width = (img.shape[:2])\n",
        "    kernel_height, kernel_width = (kernel.shape[:2])\n",
        "\n",
        "    if img_height < kernel_height or img_width < kernel_width:\n",
        "        return img\n",
        "\n",
        "    kernel_mid_x = math.ceil(kernel_width / 2)\n",
        "    kernel_mid_y = math.ceil(kernel_height / 2)\n",
        "\n",
        "    image_result = None\n",
        "\n",
        "    if mode == 'valid':\n",
        "\n",
        "        new_image_height = img_height - kernel_height - 1\n",
        "        new_image_width = img_width - kernel_width - 1\n",
        "        image_result = np.zeros((new_image_height, new_image_width))\n",
        "\n",
        "        for i in range(kernel_mid_y - 1, img_height - kernel_mid_y + 1):\n",
        "            for j in range(kernel_mid_x - 1, img_width - kernel_mid_x + 1):\n",
        "                image_result[i - kernel_mid_y - 1][j - kernel_mid_x - 1] = dot_product(j, i, img_height, img_width,\n",
        "                                                                                  kernel_height, kernel_width, img, kernel,\n",
        "                                                                                  mode)\n",
        "\n",
        "    elif mode == 'same':\n",
        "        image_result = np.zeros((img_height, img_width))\n",
        "\n",
        "        for i in range(img_height):\n",
        "            for j in range(img_width):\n",
        "                image_result[i - kernel_mid_y - 1][j - kernel_mid_x - 1] = dot_product(j, i, img_height, img_width,\n",
        "                                                                                  kernel_height, kernel_width, img, kernel,\n",
        "                                                                                  mode)\n",
        "\n",
        "    elif mode == 'full':\n",
        "        image_result = np.zeros((img_height + kernel_height - 1, img_width + kernel_width - 1))\n",
        "\n",
        "        for i in range(img_height):\n",
        "            for j in range(img_width):\n",
        "                image_result[i - kernel_mid_y - 1][j - kernel_mid_x - 1] = dot_product(j, i, img_height, img_width,\n",
        "                                                                                  kernel_height, kernel_width, img, kernel,\n",
        "                                                                                  mode)\n",
        "\n",
        "    return image_result\n",
        "\n",
        "\n",
        "def dot_product(img_x, img_y, img_height, img_width, kernel_height, kernel_width, img, kernel, mode):\n",
        "    kernel_mid_x = math.ceil(kernel_width / 2)\n",
        "    kernel_mid_y = math.ceil(kernel_height / 2)\n",
        "    my_sum = 0\n",
        "\n",
        "    if mode == 'valid':\n",
        "        for i in range(math.ceil(-kernel_height / 2), math.ceil(kernel_height / 2)):\n",
        "            for j in range(math.ceil(-kernel_width / 2), math.ceil(kernel_width / 2)):\n",
        "                my_sum += img[img_y + i][img_x + j] * kernel[kernel_mid_y - 1 + i][kernel_mid_x - 1 + j]\n",
        "\n",
        "    elif mode == 'same':\n",
        "        for i in range(math.ceil(-kernel_height / 2), math.ceil(kernel_height / 2)):\n",
        "            for j in range(math.ceil(-kernel_width / 2), math.ceil(kernel_width / 2)):\n",
        "                if (((img_x + j) < img_width and (img_x + j) >= 0 and (img_y + i) < img_height and (img_y + i) >= 0)):\n",
        "                    my_sum += img[img_y + i][img_x + j] * kernel[kernel_mid_y - 1 + i][kernel_mid_x - 1 + j]\n",
        "\n",
        "    elif mode == 'full':\n",
        "        for i in range(math.ceil(-kernel_height / 2), math.ceil(kernel_height / 2)):\n",
        "            for j in range(math.ceil(-kernel_width / 2), math.ceil(kernel_width / 2)):\n",
        "                if ((img_y + i) < img_height and (img_y + i) >= 0) and ((img_x + j) < img_width and (img_x + j) >= 0):\n",
        "                    my_sum += img[img_y + i][img_x + j] * kernel[kernel_mid_y - 1 + i][kernel_mid_x - 1 + j]\n",
        "\n",
        "    return my_sum\n",
        "\n",
        "\n",
        "def apply_convolution(img, kernel, mode='valid'):\n",
        "    new_kernel = np.fliplr(np.flipud(kernel))\n",
        "    return apply_correlation(img, new_kernel, mode)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 0,
      "metadata": {
        "colab": {},
        "colab_type": "code",
        "id": "4hRWOoqzUHNF"
      },
      "outputs": [],
      "source": [
        "########################################################################\n",
        "# Histogram Equalization \n",
        "# https://www.researchgate.net/publication/283727396_Image_enhancement_by_Histogram_equalization\n",
        "########################################################################\n",
        "\n",
        "def apply_histogram_equalization(img):\n",
        "    new_img = img.copy().flatten()\n",
        "    hist_result = np.histogram(new_img, 256, density=True)\n",
        "    cdf = 255 * hist_result[0].cumsum() / hist_result[0].cumsum()[-1]\n",
        "    equalized_img = np.interp(new_img, hist_result[1][:-1], cdf).reshape(img.shape)\n",
        "    return equalized_img"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 0,
      "metadata": {
        "colab": {},
        "colab_type": "code",
        "id": "RslSdZqXb5Gs"
      },
      "outputs": [],
      "source": [
        "########################################################################\n",
        "# Bilateral Filter\n",
        "# https://www.cs.jhu.edu/~misha/ReadingSeminar/Papers/Tomasi98.pdf\n",
        "# https://pdfs.semanticscholar.org/3796/3f1672c4f53f57a1774c381aaf86695f2087.pdf\n",
        "########################################################################\n",
        "\n",
        "\n",
        "def bilateral_filter(orgn, filtered_img, x, y, d, i_sigma, s_sigma):\n",
        "\n",
        "    Wp = 0\n",
        "    i_filter = 0\n",
        "\n",
        "    for i in range(d):\n",
        "        for j in range(d):\n",
        "            x_neigh = int(x - (d / 2 - i))\n",
        "            y_neigh = int(y - (d / 2 - j))\n",
        "\n",
        "            x_len = len(orgn)\n",
        "            y_len = len(orgn[0])\n",
        "\n",
        "            if x_neigh >= x_len:\n",
        "                x_neigh -= x_len\n",
        "            if y_neigh >= y_len:\n",
        "                y_neigh -= y_len\n",
        "\n",
        "            gi = (1.0 / (2 * math.pi * (i_sigma ** 2))) \\\n",
        "                 * math.exp(- (orgn[x_neigh][y_neigh] - orgn[x][y] ** 2) / (2 * i_sigma ** 2))\n",
        "            gs = (1.0 / (2 * math.pi * (s_sigma ** 2))) \\\n",
        "                 * math.exp(- (np.sqrt((x_neigh - x)**2 + (y_neigh - y)**2) ** 2) / (2 * s_sigma ** 2))\n",
        "\n",
        "            i_filter += orgn[x_neigh][y_neigh] * (gi * gs)\n",
        "            Wp = Wp + (gi * gs)\n",
        "\n",
        "    new = int(round(i_filter / Wp))\n",
        "    filtered_img[x][y] = new\n",
        "\n",
        "\n",
        "def apply_bilateral_filter(img, d, i_sigma, s_sigma):\n",
        "    filtered_image = np.zeros(img.shape)\n",
        "\n",
        "    for i in range(len(img)):\n",
        "        for j in range(len(img[0])):\n",
        "            bilateral_filter(img, filtered_image, i, j, d, i_sigma, s_sigma)\n",
        "\n",
        "    return filtered_image\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 0,
      "metadata": {
        "colab": {},
        "colab_type": "code",
        "id": "pMRfostZI2IU"
      },
      "outputs": [],
      "source": [
        "########################################################################\n",
        "# Thresholding\n",
        "# https://www.researchgate.net/publication/220494200_Adaptive_Thresholding_using_the_Integral_Image\n",
        "########################################################################\n",
        "\n",
        "def apply_threshold(img, s, T = 0.15):\n",
        "    result_img = np.zeros(img.shape)\n",
        "    img_height = img.shape[0]\n",
        "    img_width = img.shape[1]\n",
        "\n",
        "    for i in range(img_height):\n",
        "        y1 = i - int(int(max(img_height, img_width) / 8) / 4)\n",
        "        y2 = i + int(int(max(img_height, img_width) / 8) / 4)\n",
        "\n",
        "        if y1 < 0:\n",
        "            y1 = 0\n",
        "        if y2 >= img_height:\n",
        "            y2 = img_height - 1\n",
        "\n",
        "        for j in range(img_width):\n",
        "            x1 = j - int(int(max(img_height, img_width) / 8) / 4)\n",
        "            x2 = j + int(int(max(img_height, img_width) / 8) / 4)\n",
        "\n",
        "            if x1 < 0:\n",
        "                x1 = 0\n",
        "            if x2 >= img_width:\n",
        "                x2 = img_width - 1\n",
        "            count = (x2 - x1) * (y2 - y1)\n",
        "\n",
        "            V = ((s[y2][x2]-s[y2][x1]-s[y1][x2]+s[y1][x1]) * (1.0 - T))\n",
        "            \n",
        "            if (int)(img[i][j][0] * count) < (int)(V[0]):\n",
        "                result_img[i][j] = 0\n",
        "            else:\n",
        "                result_img[i][j] = 255\n",
        "    return result_img\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 0,
      "metadata": {
        "colab": {},
        "colab_type": "code",
        "id": "pkybIsMyRLAg"
      },
      "outputs": [],
      "source": [
        "########################################################################\n",
        "# Skew Detection\n",
        "# https://www.pyimagesearch.com/2017/02/20/text-skew-correction-opencv-python/\n",
        "########################################################################\n",
        "\n",
        "def apply_skew_detection(img):\n",
        "    image = cv2.imread(img)\n",
        "    gray_image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)\n",
        "    inverted_image = cv2.bitwise_not(gray_image)\n",
        "\n",
        "    image_threshold = cv2.threshold(inverted_image, 0, 255,\n",
        "                           cv2.THRESH_BINARY | cv2.THRESH_OTSU)[1]\n",
        "\n",
        "    skew_angle = cv2.minAreaRect(np.column_stack(np.where(image_threshold > 0)))[-1]\n",
        "\n",
        "    if skew_angle >= -45:\n",
        "        skew_angle = -skew_angle\n",
        "    else:\n",
        "        skew_angle = -(90 + skew_angle)\n",
        "\n",
        "    # de-skew the image\n",
        "    (height, width) = image.shape[:2]\n",
        "    rotated = cv2.warpAffine(image, \n",
        "                             cv2.getRotationMatrix2D((width // 2, height // 2), skew_angle, 1.0), \n",
        "                             (width, height),\n",
        "                             flags=cv2.INTER_CUBIC, \n",
        "                             borderMode=cv2.BORDER_REPLICATE)\n",
        "\n",
        "    return rotated\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 0,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 269
        },
        "colab_type": "code",
        "id": "e2ymYGolMVER",
        "outputId": "cb51488d-7140-433a-8499-4d3d56347b5f"
      },
      "outputs": [
        {
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAANEAAAD8CAYAAADpCEEHAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "########################################################################\n",
        "# PRE-PROCESS THE IMAGE\n",
        "########################################################################\n",
        "\n",
        "def pre_process_img(image_path):\n",
        "    img = cv2.imread(image_path)\n",
        "\n",
        "    img = convert_to_greyscale(img)\n",
        "    cv2.imwrite('/content/drive/My Drive/grey_img.jpg', img)\n",
        "\n",
        "    img = apply_convolution('/content/drive/My Drive/grey_img.jpg', cv2.getGaussianKernel(3, 3), 'same')\n",
        "    cv2.imwrite('/content/drive/My Drive/blur_img.jpg', img)\n",
        "\n",
        "    img = cv2.imread('/content/drive/My Drive/blur_img.jpg', cv2.IMREAD_GRAYSCALE)\n",
        "    img = apply_histogram_equalization(img)\n",
        "    cv2.imwrite('/content/drive/My Drive/equalized.jpg', img)\n",
        "\n",
        "    new_img = apply_bilateral_filter(img, 8, 14.0, 18.0)\n",
        "    cv2.imwrite('/content/drive/My Drive/bilateral_filtered.jpg', img)\n",
        "\n",
        "    img = cv2.imread('/content/drive/My Drive/bilateral_filtered.jpg')\n",
        "    thresh = apply_threshold(img, cv2.integral(img))\n",
        "    cv2.imwrite('/content/drive/My Drive/threshold.jpg', thresh)\n",
        "\n",
        "    img = apply_skew_detection('/content/drive/My Drive/threshold.jpg')\n",
        "    cv2.imwrite('/content/drive/My Drive/final_img.jpg', img)\n",
        "\n",
        "    return img\n",
        "\n",
        "\n",
        "img_path = '/content/drive/My Drive/img3.jpeg'\n",
        "processed_img = pre_process_img(img_path)\n",
        "plt.imshow(processed_img, cmap = plt.get_cmap('gray'))\n",
        "plt.show()"
      ]
    }
  ]
}