jpeg_compression.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 226,
   "metadata": {
    "collapsed": true,
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "import torch\n",
    "import numpy as np\n",
    "from torchvision.datasets import CIFAR10\n",
    "from torch.utils.data import DataLoader\n",
    "import torchvision.transforms as transforms\n",
    "import os\n",
    "from scipy.fft import dctn, idctn\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "torch.set_printoptions(threshold=10_000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 165,
   "outputs": [],
   "source": [
    "download_path = os.path.join(os.getcwd(), 'cifar10')\n",
    "\n",
    "if not os.path.exists(download_path):\n",
    "    os.makedirs(download_path)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "outputs": [],
   "source": [
    "transform = transforms.Compose([transforms.ToTensor()])\n",
    "\n",
    "train_dataset = CIFAR10(download_path, transform=transform, download=True)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n",
     "is_executing": true
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 339,
   "outputs": [
    {
     "data": {
      "text/plain": "array([[[ 59,  62,  63],\n        [ 43,  46,  45],\n        [ 50,  48,  43],\n        ...,\n        [158, 132, 108],\n        [152, 125, 102],\n        [148, 124, 103]],\n\n       [[ 16,  20,  20],\n        [  0,   0,   0],\n        [ 18,   8,   0],\n        ...,\n        [123,  88,  55],\n        [119,  83,  50],\n        [122,  87,  57]],\n\n       [[ 25,  24,  21],\n        [ 16,   7,   0],\n        [ 49,  27,   8],\n        ...,\n        [118,  84,  50],\n        [120,  84,  50],\n        [109,  73,  42]],\n\n       ...,\n\n       [[208, 170,  96],\n        [201, 153,  34],\n        [198, 161,  26],\n        ...,\n        [160, 133,  70],\n        [ 56,  31,   7],\n        [ 53,  34,  20]],\n\n       [[180, 139,  96],\n        [173, 123,  42],\n        [186, 144,  30],\n        ...,\n        [184, 148,  94],\n        [ 97,  62,  34],\n        [ 83,  53,  34]],\n\n       [[177, 144, 116],\n        [168, 129,  94],\n        [179, 142,  87],\n        ...,\n        [216, 184, 140],\n        [151, 118,  84],\n        [123,  92,  72]]], dtype=uint8)"
     },
     "execution_count": 339,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_dataset.data[0]"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 167,
   "outputs": [
    {
     "data": {
      "text/plain": "torch.Size([3, 32, 32])"
     },
     "execution_count": 167,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_dataset[0][0].shape"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 168,
   "outputs": [],
   "source": [
    "rgb_image = train_dataset[0][0]"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 169,
   "outputs": [],
   "source": [
    "def rgb_to_ycbcr(image: torch.Tensor) -> torch.Tensor:\n",
    "    r\"\"\"Convert an RGB image to YCbCr.\n",
    "\n",
    "    Args:\n",
    "        image (torch.Tensor): RGB Image to be converted to YCbCr.\n",
    "\n",
    "    Returns:\n",
    "        torch.Tensor: YCbCr version of the image.\n",
    "    \"\"\"\n",
    "\n",
    "    if not torch.is_tensor(image):\n",
    "        raise TypeError(\"Input type is not a torch.Tensor. Got {}\".format(\n",
    "            type(image)))\n",
    "\n",
    "    if len(image.shape) < 3 or image.shape[-3] != 3:\n",
    "        raise ValueError(\"Input size must have a shape of (*, 3, H, W). Got {}\"\n",
    "                         .format(image.shape))\n",
    "\n",
    "    r: torch.Tensor = image[..., 0, :, :]\n",
    "    g: torch.Tensor = image[..., 1, :, :]\n",
    "    b: torch.Tensor = image[..., 2, :, :]\n",
    "\n",
    "    delta = .5\n",
    "    y: torch.Tensor = .299 * r + .587 * g + .114 * b\n",
    "    cb: torch.Tensor = -0.168736 * r - 0.331264 * g + .5 * b + 128\n",
    "    cr: torch.Tensor = .5 * r - .418688 * g - .081312 * b + 128\n",
    "    return torch.stack((y, cb, cr), -3)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 170,
   "outputs": [],
   "source": [
    "def ycbcr_to_rgb(image: torch.Tensor) -> torch.Tensor:\n",
    "\n",
    "    r\"\"\"Convert an YCbCr image to RGB.\n",
    "\n",
    "     Args:\n",
    "         image (torch.Tensor): YCbCr Image to be converted to RGB.\n",
    "\n",
    "     Returns:\n",
    "         torch.Tensor: RGB version of the image.\n",
    "     \"\"\"\n",
    "\n",
    "    if not torch.is_tensor(image):\n",
    "        raise TypeError(\"Input type is not a torch.Tensor. Got {}\".format(\n",
    "            type(image)))\n",
    "\n",
    "    if len(image.shape) < 3 or image.shape[-3] != 3:\n",
    "        raise ValueError(\"Input size must have a shape of (*, 3, H, W). Got {}\"\n",
    "                         .format(image.shape))\n",
    "\n",
    "    Y: torch.Tensor = image[..., 0, :, :]\n",
    "    Cb: torch.Tensor = image[..., 1, :, :] - 128\n",
    "    Cr: torch.Tensor = image[..., 2, :, :] - 128\n",
    "\n",
    "    R: torch.Tensor = Y + 1.40210 * Cr\n",
    "    G: torch.Tensor = Y - 0.34414 * Cb - .71414 * Cr\n",
    "    B: torch.Tensor = Y + 1.7718 * Cb\n",
    "    return torch.stack((R, G, B), -3)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 254,
   "outputs": [],
   "source": [
    "def blockwise_dct(image: torch.Tensor, block_size: tuple[int, int]=(8, 8)):\n",
    "    # Get the image dimensions\n",
    "    image = image.numpy()\n",
    "    height, width = image.shape\n",
    "    block_height, block_width = block_size\n",
    "\n",
    "    # Prepare an array to hold the DCT coefficients\n",
    "    dct_blocks = np.zeros_like(image, dtype=np.float32)\n",
    "\n",
    "    # Process each block\n",
    "    for i in range(0, height, block_height):\n",
    "        for j in range(0, width, block_width):\n",
    "            # Extract the block from the image\n",
    "            block = image[i:i+block_height, j:j+block_width]\n",
    "\n",
    "            # Apply the 2D DCT to the block\n",
    "            dct_block = dctn(block, norm='ortho')\n",
    "\n",
    "            # Store the DCT coefficients\n",
    "            dct_blocks[i:i+block_height, j:j+block_width] = dct_block\n",
    "\n",
    "    return torch.from_numpy(dct_blocks)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 279,
   "outputs": [],
   "source": [
    "def blockwise_quantize_luminance(dct: torch.Tensor, mode = 'l', alpha: int = 1, block_size = (8, 8)) -> torch.Tensor:\n",
    "    luminance_quantization_matrix = torch.tensor([\n",
    "        [16, 11, 10, 16, 24, 40, 51, 61],\n",
    "        [12, 12, 14, 19, 26, 58, 60, 55],\n",
    "        [14, 13, 16, 24, 40, 57, 69, 56],\n",
    "        [14, 17, 22, 29, 51, 87, 80, 62],\n",
    "        [18, 22, 37, 56, 68, 109, 103, 77],\n",
    "        [24, 35, 55, 64, 81, 104, 113, 92],\n",
    "        [49, 64, 78, 87, 103, 121, 120, 101],\n",
    "        [72, 92, 95, 98, 112, 100, 103, 99]\n",
    "    ])\n",
    "\n",
    "    chrominance_quantization_matrix = torch.tensor([\n",
    "        [17, 18, 24, 47, 99, 99, 99, 99],\n",
    "        [18, 21, 26, 66, 99, 99, 99, 99],\n",
    "        [24, 26, 56, 99, 99, 99, 99, 99],\n",
    "        [47, 66, 99, 99, 99, 99, 99, 99],\n",
    "        [99, 99, 99, 99, 99, 99, 99, 99],\n",
    "        [99, 99, 99, 99, 99, 99, 99, 99],\n",
    "        [99, 99, 99, 99, 99, 99, 99, 99],\n",
    "        [99, 99, 99, 99, 99, 99, 99, 99]\n",
    "    ])\n",
    "\n",
    "    quantization_matrix = luminance_quantization_matrix if mode == 'l' else chrominance_quantization_matrix\n",
    "\n",
    "\n",
    "    height, width = dct.shape\n",
    "    block_height_num, block_width_num = height//block_size[0], width//block_size[1]\n",
    "\n",
    "\n",
    "    quantization_matrix *= alpha\n",
    "    quantization_matrix = quantization_matrix.repeat(block_height_num, block_width_num)\n",
    "\n",
    "    quantisation_blocks = torch.zeros((height, width), dtype=torch.float32)\n",
    "\n",
    "    return torch.round(torch.div(dct, quantization_matrix)).to(torch.int8)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 284,
   "outputs": [],
   "source": [
    "def zigzag_order(block: torch.Tensor) -> torch.Tensor:\n",
    "    # Zigzag index pattern for an 8x8 block\n",
    "    zigzag_indices = torch.tensor([\n",
    "        0, 1, 8, 16, 9, 2, 3, 10, 17, 24, 32, 25, 18, 11, 4, 5,\n",
    "        12, 19, 26, 33, 40, 48, 41, 34, 27, 20, 13, 6, 7, 14, 21, 28,\n",
    "        35, 42, 49, 56, 57, 50, 43, 36, 29, 22, 15, 23, 30, 37, 44, 51,\n",
    "        58, 59, 52, 45, 38, 31, 39, 46, 53, 60, 61, 54, 47, 55, 62, 63\n",
    "    ])\n",
    "\n",
    "    # Flatten the block and reorder it using the zigzag indices\n",
    "    flat_block = block.flatten()\n",
    "    zigzag_block = flat_block[zigzag_indices]\n",
    "\n",
    "    return zigzag_block"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 328,
   "outputs": [],
   "source": [
    "def blockwise_zigzag_order(image: torch.Tensor, block_size: tuple[int, int] = (8, 8)) -> torch.Tensor:\n",
    "\n",
    "    img_height, img_width = image.shape\n",
    "    block_height, block_width = block_size\n",
    "    block_height_num, block_width_num = img_height//block_height, img_width//block_width\n",
    "    zigzag_length = block_size[0] * block_size[1]\n",
    "\n",
    "    zigzag_patches = torch.zeros((block_height_num, block_width_num, zigzag_length), dtype = torch.int8)\n",
    "\n",
    "\n",
    "    for i in range(0, img_height, block_height):\n",
    "        for j in range(0, img_width, block_width):\n",
    "            block = image[i:i+block_height, j:j+block_width]\n",
    "            zigzag_block = zigzag_order(block)\n",
    "            zigzag_patches[i//block_height, j//block_width, :] = zigzag_block\n",
    "\n",
    "    return zigzag_patches"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "# Checking the code"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 329,
   "outputs": [],
   "source": [
    "ycbcr_image = rgb_to_ycbcr(rgb_image * 255)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 330,
   "outputs": [],
   "source": [
    "luminance_patch = ycbcr_image[..., 0, 0:8, 0:8]"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 331,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([32, 32])\n"
     ]
    },
    {
     "data": {
      "text/plain": "tensor([[[ 37, -13,  -7,   3,  -7,  -4,   1,   2,   1,   2,   1,   1,   3,   1,\n            0,   0,   0,   0,   0,   1,   1,   1,   1,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0],\n         [ 42,   4,   7,   0,  -5,   1,   0,   1,   5,   3,   2,   0,   0,   1,\n           -1,   0,   0,   0,   0,   0,   2,   0,   0,   0,   1,   1,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0],\n         [ 44,  -2,   8,   0,  -2,   4,   1,  -4,   4,  -1,   4,  -2,   3,   0,\n            0,   0,   1,   0,  -1,   1,   0,   0,  -1,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0],\n         [ 45,   8,   9,   6,  -5,  -2,   0,  -2,  -2,  -1,   3,   3,   1,   0,\n           -1,   0,   1,  -1,   0,  -1,   2,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0]],\n\n        [[ 50,  12,   3,  -2,  -5,  -1,  -1,   2,   0,  -2,  -2,   1,   1,   0,\n            1,   0,  -1,  -1,   0,  -1,   0,   0,  -1,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0],\n         [ 55, -11, -23,  -2,  16,   5,  -3,   3,   5,   2,   0,  -3,   2,   4,\n           -1,   1,   1,  -3,  -3,   1,   0,   0,   0,   0,  -1,  -1,   0,   0,\n            0,   0,  -1,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   1,\n            0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0],\n         [ 72,   8, -36,   3,   2,  -1,   2,  -1,   3,  -1,   2,   4,   4,   1,\n           -1,   0,  -1,  -2,   0,   1,   1,   0,  -1,   0,  -1,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0],\n         [ 51,   3,  -5,   2,  -7,   2,   1,  -5,   4,  -1,   0,  -1,   3,  -3,\n            1,   0,  -1,   1,   0,   0,   0,   0,   0,   0,   0,   1,  -1,   0,\n            0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0]],\n\n        [[ 54,  -6,  -3,  -5,   5,   8,  -2,   0,   9,   0,   2,   0,  -5,  -2,\n            3,   0,   1,   1,   0,  -2,  -1,   0,   1,   1,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0],\n         [ 80,   7,  11,   1,  -4,   2,   6,   8,  -8,  -5,  -3,  -1,  -1,  -1,\n           -1,  -2,  -2,  -4,   1,   1,   0,   0,   1,   0,  -2,  -1,  -1,   0,\n            0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0],\n         [ 73,  -6,  23,   3,  -2,   7,  -1,   5,  11,   2,   2,   4,  -2,  -2,\n            0,   0,   1,  -1,   1,  -1,   1,   0,  -1,   0,  -1,   1,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0],\n         [ 58,   5,   2,   4,  12,  13,   2,   2,  -3,   1,  -2,   0,  -4,   5,\n            3,   0,   1,  -2,  -1,   1,   0,   0,   0,   0,   1,  -1,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0]],\n\n        [[ 65,   4, -12,   2,   1,   5,   2,   3,  -8,   0,   1,   1,  -2,   0,\n            1,   1,   1,   1,  -2,   0,  -1,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0],\n         [ 59,   1,   6,  10,  -3,  -1,   3,  -2,   0,   1,   3,  -1,   0,   1,\n            0,   0,   0,   0,  -1,   0,  -1,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0],\n         [ 51,   4,  -2,  -1,  -1,  -2,  -1,   0,   4,  -3,   0,   2,   0,   0,\n            1,   0,   0,   1,  -2,   1,   0,   0,   0,  -1,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0],\n         [ 53,  -3,   8,   1,  -3,  -5,  10,   1,  -7,  -6,   2,   3,   4,  -4,\n           -3,  -1,  -1,  -2,   1,   1,  -1,   0,   0,  -1,  -1,   0,   1,   1,\n            0,  -1,   1,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n            0,   0,   0,   0,   0,   0,   0,   0]]], dtype=torch.int8)"
     },
     "execution_count": 331,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dct_values = blockwise_dct(ycbcr_image[0])\n",
    "quantized_blocks = blockwise_quantize_luminance(dct_values)\n",
    "print(quantized_blocks.shape)\n",
    "# quantized_blocks[0:8, 0:8]\n",
    "blockwise_zigzag_order(quantized_blocks)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n"
     ]
    },
    {
     "data": {
      "text/plain": "<Figure size 640x480 with 1 Axes>",
      "image/png": "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"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "recover = ycbcr_to_rgb(rgb_to_ycbcr(rgb_image*255))\n",
    "plt.imshow(recover.permute(1, 2, 0) / 255)\n",
    "plt.show()"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "outputs": [
    {
     "data": {
      "text/plain": "<Figure size 640x480 with 1 Axes>",
      "image/png": "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"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(rgb_image.permute(1, 2, 0))\n",
    "plt.show()"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "# Old"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "outputs": [],
   "source": [
    "def _patch_to_dct(luminance: torch.Tensor) -> torch.Tensor:\n",
    "    pi = torch.tensor(torch.pi)\n",
    "\n",
    "    dct_luminance_patch = torch.zeros(*luminance.shape)\n",
    "\n",
    "    for u in range(0, 8):\n",
    "        for v in range(0,8):\n",
    "            Cu = torch.sqrt(torch.tensor(0.5)) if u == 0 else torch.tensor(1)\n",
    "            Cv = torch.sqrt(torch.tensor(0.5)) if v == 0 else torch.tensor(1)\n",
    "\n",
    "            values = torch.zeros(8,8)\n",
    "            for x in range(0, 8):\n",
    "                for y in range(0, 8):\n",
    "                    u_, v_, x_, y_ = torch.tensor(u), torch.tensor(v), torch.tensor(x), torch.tensor(y)\n",
    "                    cosine = torch.cos(((2*x_+1)*u_*pi)/16)*torch.cos(((2*y_+1)*v_*pi)/16)\n",
    "                    values[x, y] = luminance_patch[x, y] * cosine\n",
    "\n",
    "            dct_luminance_patch[u, v] = 0.25 * Cu * Cv * torch.sum(values)\n",
    "\n",
    "    return dct_luminance_patch"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "outputs": [],
   "source": [
    "def quantize_luminance(dct: torch.Tensor, alpha: int = 1) -> torch.Tensor:\n",
    "    luminance_quantization_matrix = torch.tensor([\n",
    "        [16, 11, 10, 16, 24, 40, 51, 61],\n",
    "        [12, 12, 14, 19, 26, 58, 60, 55],\n",
    "        [14, 13, 16, 24, 40, 57, 69, 56],\n",
    "        [14, 17, 22, 29, 51, 87, 80, 62],\n",
    "        [18, 22, 37, 56, 68, 109, 103, 77],\n",
    "        [24, 35, 55, 64, 81, 104, 113, 92],\n",
    "        [49, 64, 78, 87, 103, 121, 120, 101],\n",
    "        [72, 92, 95, 98, 112, 100, 103, 99]\n",
    "    ])\n",
    "\n",
    "    luminance_quantization_matrix *= alpha\n",
    "\n",
    "    return torch.round(torch.div(dct, luminance_quantization_matrix)).to(torch.int8)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "outputs": [],
   "source": [
    "def compute_dct(image_patch):\n",
    "    # Convert from PyTorch tensor to NumPy array if necessary\n",
    "    if isinstance(image_patch, torch.Tensor):\n",
    "        image_patch = image_patch.numpy()\n",
    "\n",
    "    # Apply DCT in both dimensions using dctn for n-dimensional DCT\n",
    "    dct_2d = dctn(image_patch, norm='ortho')\n",
    "\n",
    "    # Convert back to PyTorch tensor from NumPy array\n",
    "    return torch.from_numpy(dct_2d)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 270,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor([[ 37, -13,  -4,   1,   0,   0,   0,   0],\n",
      "        [ -7,  -7,   2,   1,   0,   0,   0,   0],\n",
      "        [  3,   1,   3,   0,   0,   0,   0,   0],\n",
      "        [  2,   1,   0,   0,   0,   0,   0,   0],\n",
      "        [  1,   1,   0,   0,   0,   0,   0,   0],\n",
      "        [  1,   1,   0,   0,   0,   0,   0,   0],\n",
      "        [  1,   0,   0,   0,   0,   0,   0,   0],\n",
      "        [  0,   0,   0,   0,   0,   0,   0,   0]], dtype=torch.int8)\n",
      "tensor([[ 37, -13,  -4,   1,   0,   0,   0,   0],\n",
      "        [ -7,  -7,   2,   1,   0,   0,   0,   0],\n",
      "        [  3,   1,   3,   0,   0,   0,   0,   0],\n",
      "        [  2,   1,   0,   0,   0,   0,   0,   0],\n",
      "        [  1,   1,   0,   0,   0,   0,   0,   0],\n",
      "        [  1,   1,   0,   0,   0,   0,   0,   0],\n",
      "        [  1,   0,   0,   0,   0,   0,   0,   0],\n",
      "        [  0,   0,   0,   0,   0,   0,   0,   0]], dtype=torch.int8)\n",
      "tensor([[ 37, -13,  -4,   1,   0,   0,   0,   0],\n",
      "        [ -7,  -7,   2,   1,   0,   0,   0,   0],\n",
      "        [  3,   1,   3,   0,   0,   0,   0,   0],\n",
      "        [  2,   1,   0,   0,   0,   0,   0,   0],\n",
      "        [  1,   1,   0,   0,   0,   0,   0,   0],\n",
      "        [  1,   1,   0,   0,   0,   0,   0,   0],\n",
      "        [  1,   0,   0,   0,   0,   0,   0,   0],\n",
      "        [  0,   0,   0,   0,   0,   0,   0,   0]], dtype=torch.int8)\n",
      "tensor([[True, True, True, True, True, True, True, True],\n",
      "        [True, True, True, True, True, True, True, True],\n",
      "        [True, True, True, True, True, True, True, True],\n",
      "        [True, True, True, True, True, True, True, True],\n",
      "        [True, True, True, True, True, True, True, True],\n",
      "        [True, True, True, True, True, True, True, True],\n",
      "        [True, True, True, True, True, True, True, True],\n",
      "        [True, True, True, True, True, True, True, True]])\n",
      "tensor([[True, True, True, True, True, True, True, True],\n",
      "        [True, True, True, True, True, True, True, True],\n",
      "        [True, True, True, True, True, True, True, True],\n",
      "        [True, True, True, True, True, True, True, True],\n",
      "        [True, True, True, True, True, True, True, True],\n",
      "        [True, True, True, True, True, True, True, True],\n",
      "        [True, True, True, True, True, True, True, True],\n",
      "        [True, True, True, True, True, True, True, True]])\n"
     ]
    }
   ],
   "source": [
    "dct = compute_dct(luminance_patch)\n",
    "dct_2 = _patch_to_dct(luminance_patch)\n",
    "quant_1 = quantize_luminance(dct)\n",
    "quant_2 = quantize_luminance(dct_2)\n",
    "\n",
    "dct_values = blockwise_dct(ycbcr_image[0])\n",
    "quantized_blocks = blockwise_quantize_luminance(dct_values)\n",
    "quant_3 = quantized_blocks[0:8, 0:8]\n",
    "\n",
    "print(quant_1)\n",
    "print(quant_2)\n",
    "print(quant_3)\n",
    "print(quant_1 == quant_3)\n",
    "print(quant_2 == quant_3)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 285,
   "outputs": [
    {
     "data": {
      "text/plain": "tensor([ 37, -13,  -7,   3,  -7,  -4,   1,   2,   1,   2,   1,   1,   3,   1,\n          0,   0,   0,   0,   0,   1,   1,   1,   1,   0,   0,   0,   0,   0,\n          0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n          0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n          0,   0,   0,   0,   0,   0,   0,   0], dtype=torch.int8)"
     },
     "execution_count": 285,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "zigzag_order(quant_1)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "# Cifar source code"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "outputs": [],
   "source": [
    "import os.path\n",
    "import pickle\n",
    "from typing import Any, Callable, Optional, Tuple\n",
    "\n",
    "import numpy as np\n",
    "from PIL import Image\n",
    "\n",
    "from .utils import check_integrity, download_and_extract_archive\n",
    "from .vision import VisionDataset\n",
    "\n",
    "\n",
    "class CIFAR10(VisionDataset):\n",
    "    \"\"\"`CIFAR10 <https://www.cs.toronto.edu/~kriz/cifar.html>`_ Dataset.\n",
    "\n",
    "    Args:\n",
    "        root (string): Root directory of dataset where directory\n",
    "            ``cifar-10-batches-py`` exists or will be saved to if download is set to True.\n",
    "        train (bool, optional): If True, creates dataset from training set, otherwise\n",
    "            creates from test set.\n",
    "        transform (callable, optional): A function/transform that takes in an PIL image\n",
    "            and returns a transformed version. E.g, ``transforms.RandomCrop``\n",
    "        target_transform (callable, optional): A function/transform that takes in the\n",
    "            target and transforms it.\n",
    "        download (bool, optional): If true, downloads the dataset from the internet and\n",
    "            puts it in root directory. If dataset is already downloaded, it is not\n",
    "            downloaded again.\n",
    "\n",
    "    \"\"\"\n",
    "\n",
    "    base_folder = \"cifar-10-batches-py\"\n",
    "    url = \"https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz\"\n",
    "    filename = \"cifar-10-python.tar.gz\"\n",
    "    tgz_md5 = \"c58f30108f718f92721af3b95e74349a\"\n",
    "    train_list = [\n",
    "        [\"data_batch_1\", \"c99cafc152244af753f735de768cd75f\"],\n",
    "        [\"data_batch_2\", \"d4bba439e000b95fd0a9bffe97cbabec\"],\n",
    "        [\"data_batch_3\", \"54ebc095f3ab1f0389bbae665268c751\"],\n",
    "        [\"data_batch_4\", \"634d18415352ddfa80567beed471001a\"],\n",
    "        [\"data_batch_5\", \"482c414d41f54cd18b22e5b47cb7c3cb\"],\n",
    "    ]\n",
    "\n",
    "    test_list = [\n",
    "        [\"test_batch\", \"40351d587109b95175f43aff81a1287e\"],\n",
    "    ]\n",
    "    meta = {\n",
    "        \"filename\": \"batches.meta\",\n",
    "        \"key\": \"label_names\",\n",
    "        \"md5\": \"5ff9c542aee3614f3951f8cda6e48888\",\n",
    "    }\n",
    "\n",
    "    def __init__(\n",
    "            self,\n",
    "            root: str,\n",
    "            train: bool = True,\n",
    "            transform: Optional[Callable] = None,\n",
    "            target_transform: Optional[Callable] = None,\n",
    "            download: bool = False,\n",
    "    ) -> None:\n",
    "\n",
    "        super().__init__(root, transform=transform, target_transform=target_transform)\n",
    "\n",
    "        self.train = train  # training set or test set\n",
    "\n",
    "        if download:\n",
    "            self.download()\n",
    "\n",
    "        if not self._check_integrity():\n",
    "            raise RuntimeError(\"Dataset not found or corrupted. You can use download=True to download it\")\n",
    "\n",
    "        if self.train:\n",
    "            downloaded_list = self.train_list\n",
    "        else:\n",
    "            downloaded_list = self.test_list\n",
    "\n",
    "        self.data: Any = []\n",
    "        self.targets = []\n",
    "\n",
    "        # now load the picked numpy arrays\n",
    "        for file_name, checksum in downloaded_list:\n",
    "            file_path = os.path.join(self.root, self.base_folder, file_name)\n",
    "            with open(file_path, \"rb\") as f:\n",
    "                entry = pickle.load(f, encoding=\"latin1\")\n",
    "                self.data.append(entry[\"data\"])\n",
    "                if \"labels\" in entry:\n",
    "                    self.targets.extend(entry[\"labels\"])\n",
    "                else:\n",
    "                    self.targets.extend(entry[\"fine_labels\"])\n",
    "\n",
    "        self.data = np.vstack(self.data).reshape(-1, 3, 32, 32)\n",
    "        self.data = self.data.transpose((0, 2, 3, 1))  # convert to HWC\n",
    "\n",
    "        self._load_meta()\n",
    "\n",
    "    def _load_meta(self) -> None:\n",
    "        path = os.path.join(self.root, self.base_folder, self.meta[\"filename\"])\n",
    "        if not check_integrity(path, self.meta[\"md5\"]):\n",
    "            raise RuntimeError(\"Dataset metadata file not found or corrupted. You can use download=True to download it\")\n",
    "        with open(path, \"rb\") as infile:\n",
    "            data = pickle.load(infile, encoding=\"latin1\")\n",
    "            self.classes = data[self.meta[\"key\"]]\n",
    "        self.class_to_idx = {_class: i for i, _class in enumerate(self.classes)}\n",
    "\n",
    "    def __getitem__(self, index: int) -> Tuple[Any, Any]:\n",
    "        \"\"\"\n",
    "        Args:\n",
    "            index (int): Index\n",
    "\n",
    "        Returns:\n",
    "            tuple: (image, target) where target is index of the target class.\n",
    "        \"\"\"\n",
    "        img, target = self.data[index], self.targets[index]\n",
    "\n",
    "        # doing this so that it is consistent with all other datasets\n",
    "        # to return a PIL Image\n",
    "        img = Image.fromarray(img)\n",
    "\n",
    "        if self.transform is not None:\n",
    "            img = self.transform(img)\n",
    "\n",
    "        if self.target_transform is not None:\n",
    "            target = self.target_transform(target)\n",
    "\n",
    "        return img, target\n",
    "\n",
    "    def __len__(self) -> int:\n",
    "        return len(self.data)\n",
    "\n",
    "    def _check_integrity(self) -> bool:\n",
    "        for filename, md5 in self.train_list + self.test_list:\n",
    "            fpath = os.path.join(self.root, self.base_folder, filename)\n",
    "            if not check_integrity(fpath, md5):\n",
    "                return False\n",
    "        return True\n",
    "\n",
    "    def download(self) -> None:\n",
    "        if self._check_integrity():\n",
    "            print(\"Files already downloaded and verified\")\n",
    "            return\n",
    "        download_and_extract_archive(self.url, self.root, filename=self.filename, md5=self.tgz_md5)\n",
    "\n",
    "    def extra_repr(self) -> str:\n",
    "        split = \"Train\" if self.train is True else \"Test\"\n",
    "        return f\"Split: {split}\""
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}