计算机作业6（2） (1).ipynb · 胡思屹/图像去噪综合

1月18日，北京，聊聊2025如何加入技术开发？

加入 Gitee

与超过 1200万开发者一起发现、参与优秀开源项目，私有仓库也完全免费：）

文件

该仓库未声明开源许可证文件（LICENSE），使用请关注具体项目描述及其代码上游依赖。

克隆/下载

计算机作业6（2） (1).ipynb 5.19 KB

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "a79350dd-6114-4975-b3a8-0407565ac163",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "37bce153fbde417b9108fe6fb575f938",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "interactive(children=(Dropdown(description='img_list', options=(('MRI', ('img/MRI_clean.tif', 'img/MRI_noisy.t…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<function __main__.optimize_denoise(img_list)>"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from skimage import io\n",
    "import ipywidgets\n",
    "import numpy as np\n",
    "from skimage import img_as_float\n",
    "from skimage.metrics import peak_signal_noise_ratio as psnr\n",
    "from skimage.restoration import (denoise_tv_chambolle, denoise_bilateral, denoise_wavelet, denoise_nl_means)\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "def fft2_denoise(im):\n",
    "    # TODO 完成fft2去噪算法，实现输入为原图，输出为去噪后的图像\n",
    "    h,w=im.shape\n",
    "    fCoef=np.fft.fft2(im)\n",
    "    amp_im_fft=np.log(np.abs(fCoef))\n",
    "    keep = 0.1  # 每一个角落保留1/10的内容\n",
    "    mask = np.ones((h, w))\n",
    "    h_min = int(h * keep)\n",
    "    h_max = int(h * (1 - keep))  # 需要掩盖住部分的纵坐标两端\n",
    "    w_min = int(w * keep)\n",
    "    w_max = int(w * (1 - keep))  # 需要掩盖住部分的横坐标两端\n",
    "    mask[h_min:h_max, :] = 0\n",
    "    mask[:, w_min:w_max] = 0\n",
    "    fCoef_mask=fCoef*mask\n",
    "    im_denoised=np.fft.ifft2(fCoef_mask).real\n",
    "    return im_denoised\n",
    "\n",
    "def optimize_denoise(img_list):\n",
    "    # 以下虽然用了as_gray=True, 但是对于某些tif仍然无法转化为float\n",
    "    # 所以仍然要使用img_as_floatpywidgets\n",
    "    clean_img=img_as_float(io.imread(img_list[0],as_gray=True)) # 干净原图\n",
    "    noisy_img=img_as_float(io.imread(img_list[1],as_gray=True)) # 有噪音图像\n",
    "    # TODO 完成此部分代码实现题目要求功能\n",
    "    \n",
    "    # 1. 双边滤波去噪\n",
    "    bilateral_img = denoise_bilateral(noisy_img, multichannel=False) \n",
    "    # 2. 全变分去噪\n",
    "    tv_img = denoise_tv_chambolle(noisy_img)\n",
    "    # 3. 小波去噪\n",
    "    wavelet_img = denoise_wavelet(noisy_img)\n",
    "    # 4. Skimage的非局部均值去噪\n",
    "    NLM_skimg_denoise_img = denoise_nl_means(noisy_img, patch_size=9, patch_distance=5)\n",
    "    \n",
    "    # 计算所有的去噪算法的PSNR（峰值信噪比）\n",
    "    bilateral_psnr = psnr(clean_img, bilateral_img)\n",
    "    tv_psnr = psnr(clean_img, tv_img)\n",
    "    wavelet_psnr = psnr(clean_img, wavelet_img)\n",
    "    NLM_psnr = psnr(clean_img, NLM_skimg_denoise_img)\n",
    "    fft2_psnr=psnr(clean_img,fft2_denoise(noisy_img))\n",
    "    \n",
    "    print(\"bilateral PSNR:\",bilateral_psnr)\n",
    "    print(\"tv PSNR:\",tv_psnr)\n",
    "    print(\"wavelet PSNR:\",wavelet_psnr)\n",
    "    print(\"NLM PSNR:\",NLM_psnr)\n",
    "    print(\"fft2 PSNR:\",fft2_psnr)\n",
    "    \n",
    "    psnr_list=np.array([bilateral_psnr,tv_psnr,wavelet_psnr,NLM_psnr,fft2_psnr])\n",
    "    i=np.argmax(psnr_list)\n",
    "    img_list=np.array([bilateral_img,tv_img,wavelet_img,NLM_skimg_denoise_img,fft2_denoise(noisy_img)])\n",
    "    algorithm_list=np.array(['bilateral','tv','wavelet','NLM','fft2'])\n",
    "    \n",
    "    fg,ax=plt.subplots(1,3,figsize=(10,4))\n",
    "    ax[0].imshow(clean_img,cmap=\"gray\")\n",
    "    ax[0].set_title(\"clean_img\")\n",
    "    ax[1].imshow(noisy_img,cmap=\"gray\")\n",
    "    ax[1].set_title(\"noisy_img\")\n",
    "    ax[2].imshow(img_list[i],cmap=\"gray\")\n",
    "    ax[2].set_title(\"{0}:{1}\".format(algorithm_list[i],psnr_list[i]))\n",
    "    plt.tight_layout()\n",
    "    plt.show()\n",
    "\n",
    "x = ipywidgets.Dropdown(\n",
    "    options=[('MRI', (\"img/MRI_clean.tif\",\"img/MRI_noisy.tif\")),\n",
    "             ('cat', (\"img/cat.png\",\"img/cat_noisy.jpg\")),\n",
    "             ('balloon', (\"img/balloon.jpg\",\"img/balloon_noisy.jpg\")),\n",
    "             ('einstein', (\"img/einstein.jpg\",\"img/einstein_noisy.jpg\")),\n",
    "             ]\n",
    "    )\n",
    "ipywidgets.interact(optimize_denoise, img_list=x)\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "c4b5d7fe-b8cd-4f37-98c4-3afb6bbf0a22",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "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": "3.9.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}

一键复制编辑原始数据按行查看历史

提交于 2023-04-30 15:22 . 图像去噪综合应用作业

马建仓 AI 助手

尝试更多

代码解读

代码找茬

代码优化

https://gitee.com/hu-siyi/image-denoising-synthesis.git

[email protected]:hu-siyi/image-denoising-synthesis.git

hu-siyi

image-denoising-synthesis

图像去噪综合

master

胡思屹/图像去噪综合

简介

发行版

贡献者

近期动态

胡思屹/图像去噪综合 .gitee-modal { width: 500px !important; }

简介

发行版

贡献者

近期动态

搜索帮助

胡思屹/图像去噪综合