模式识别与图像处理课程实验一：图像处理实验–>> 颜色算子实验、Susan、Harris角点检测实验、 sobel边缘算子检测实验

  • 一、 实验内容
  • 二、 颜色算子实验
    • 2.1、 提取红色
    • 2.2、 提取绿色
    • 2.3、 提取蓝色
  • 三、 Susan、Harris角点检测实验
  • 3. 1、 实验程序
    • 3.1.1、Susan角点检测
    • 3.1.2、Harris角点检测
  • 四、 sobel边缘算子检测实验
    • 4.1、sobel边缘算子检
  • 五、 实验总结

一、 实验内容

要求编写一个包含颜色算子，Susan，Harris，角点，sobel边缘算子的程。

二、 颜色算子实验

2.1、 提取红色

  • 实验的程序如下
import numpy as npimport cv2 as cvimage = cv.imread("1.jpg")image = image / np.ones([1, 1, 3]).astype(np.float32)image = cv.cvtColor(image, cv.COLOR_BGR2RGB)print(image.shape)# 颜色算子# redredAdd = np.ones([1, 1, 3]).astype(np.float32)redAdd[0, 0, 0] = 1.0redAdd[0, 0, 1] = 0.5redAdd[0, 0, 2] = 0.25redSub = np.ones([1, 1, 3]).astype(np.float32)redSub[0, 0, 0] = 0.25redSub[0, 0, 1] = 0.5redSub[0, 0, 2] = 1.0image1 = np.mean(image * redAdd, 2)image2 = np.mean(image * redSub, 2) + 100imageRed = image1 / image2redMax = np.max(imageRed)redMin = np.min(imageRed)imageRed = 255 * (imageRed - redMin) / (redMax - redMin)cv.imwrite("1red.png", imageRed)

运行结果如下

  1. 实验原图
  2. 实验结果图

2.2、 提取绿色

实验的程序如下

import numpy as npimport cv2 as cvimage = cv.imread("1.jpg")image = image / np.ones([1, 1, 3]).astype(np.float32)image = cv.cvtColor(image, cv.COLOR_BGR2RGB)print(image.shape)# greengreenAdd = np.ones([1, 1, 3]).astype(np.float32)greenAdd[0, 0, 0] = 0.5greenAdd[0, 0, 1] = 1.0greenAdd[0, 0, 2] = 0.25greenSub = np.ones([1, 1, 3]).astype(np.float32)greenSub[0, 0, 0] = 0.5greenSub[0, 0, 1] = 0.25greenSub[0, 0, 2] = 1.0image1 = np.mean(image * greenAdd, 2)image2 = np.mean(image * greenSub, 2) + 100imageGreen = image1 / image2greenMax = np.max(imageGreen)greenMin = np.min(imageGreen)imageRed = 255 * (imageGreen - greenMin) / (greenMax - greenMin)cv.imwrite("1green.png", imageRed)

运行结果如下

  1. 实验原图

  2. 实验结果图

2.3、 提取蓝色

  • 实验的程序如下
import numpy as npimport cv2 as cvimage = cv.imread("1.jpg")image = image / np.ones([1, 1, 3]).astype(np.float32)image = cv.cvtColor(image, cv.COLOR_BGR2RGB)print(image.shape)# bulebuleAdd = np.ones([1, 1, 3]).astype(np.float32)buleAdd[0, 0, 0] = 0.25buleAdd[0, 0, 1] = 0.5buleAdd[0, 0, 2] = 1.0buleSub = np.ones([1, 1, 3]).astype(np.float32)buleSub[0, 0, 0] = 1.0buleSub[0, 0, 1] = 0.5buleSub[0, 0, 2] = 0.25image1 = np.mean(image * buleAdd, 2)image2 = np.mean(image * buleSub, 2) + 100imageBlue = image1 / image2blueMax = np.max(imageBlue)blueMin = np.min(imageBlue)imageBlue = 255 * (imageBlue - blueMin) / (blueMax - blueMin)cv.imwrite("1blue.png", imageBlue)

运行结果如下

  1. 实验原图

  2. 实验结果图

三、 Susan、Harris角点检测实验

3. 1、 实验程序

3.1.1、Susan角点检测

Susan角点检测程序如下

import numpy as npimport cv2 as cvimage = cv.imread("2.jpg")image = np.mean(image, 2)height = image.shape[0]width = image.shape[1]print(image.shape)#susan 算子radius = 5imageSusan = np.zeros([height, width]).astype(np.float32)for h in range(radius, height-radius):for w in range(radius, width-radius):numSmall = 0numLarge = 0numAll = 0for y in range(-radius, radius + 1):for x in range(-radius, radius+1):distance = np.sqrt(y**2 + x**2)if distance <= radius:numAll += 1if image[h + y, w + x] < image[h, w] - 27:numSmall += 1if image[h + y, w + x] > image[h, w] + 27:numLarge += 1ratio = 1.0 * numSmall / numAllratio2 = 1.0 * numLarge / numAllif ratio < 0.3:imageSusan[h, w] = 0.3 - ratioif ratio2 > 0.7:imageSusan[h, w] = ratio2 - 0.7imageMax = np.max(imageSusan)imageMin = np.min(imageSusan)imageSusan = 255*(imageSusan - imageMin)/(imageMax - imageMin)print(imageSusan.shape)cv.imwrite("2.png", imageSusan)
  • 运行结果如下

  1. 实验原图

  2. 实验结果图

3.1.2、Harris角点检测

  • Harris角点检测程序如下
import cv2 as cvimport numpy as npimport matplotlib.pyplot as plt# 读取图像img = cv.imread('3.jpg')lenna_img = cv.cvtColor(img, cv.COLOR_BGR2RGB)# 图像转换成灰度图像grayImage = cv.cvtColor(img, cv.COLOR_BGR2GRAY)grayImage = np.float32(grayImage)# Harris算子harrisImage = cv.cornerHarris(grayImage, 2, 3, 0.04)harrisImage = cv.dilate(harrisImage, None)# 设置阈值thresImage = 0.006 * harrisImage.max()img[harrisImage > thresImage] = [255, 0, 0]# 显示正常中文的标签plt.rcParams['font.sans-serif'] = ['SimHei']titles = [u'(a)原始图像', u'(b)Harris图像']images = [lenna_img, img]for i in range(2):plt.subplot(1, 2, i + 1), plt.imshow(images[i], 'gray')plt.title(titles[i])plt.xticks([]), plt.yticks([])plt.show()

运行结果如下

四、 sobel边缘算子检测实验

4.1、sobel边缘算子检

  • sobel边缘算子检程序如下
import numpy as npimport cv2image = cv2.imread("3.jpg")height = image.shape[0]width = image.shape[1]sobelResult = np.zeros([height - 2, width - 2, 1]).astype(np.float32)sobelX = np.zeros([3, 3, 1]).astype(np.float32)sobelY = np.zeros([3, 3, 1]).astype(np.float32)sobelX[0, 0, 0] = -1sobelX[1, 0, 0] = -2sobelX[2, 0, 0] = -1sobelX[0, 2, 0] = 1sobelX[1, 2, 0] = 2sobelX[2, 2, 0] = 1sobelY[0, 0, 0] = -1sobelY[0, 1, 0] = -2sobelY[0, 2, 0] = -1sobelY[2, 0, 0] = 1sobelY[2, 1, 0] = 2sobelY[2, 2, 0] = 1for h in range(0, height - 3):for w in range(0, width - 3):#求方向梯度imageIncre = image[h:h + 3, w:w + 3]gradientX = np.sum(imageIncre * sobelX)gradientY = np.sum(imageIncre * sobelY)gradient = np.sqrt(gradientX**2 + gradientY**2)sobelResult[h, w, 0] = gradientimageMax = np.max(sobelResult)imageMin = np.min(sobelResult)sobelResult = 255*(sobelResult - imageMin) / (imageMax - imageMin)cv2.imwrite("3.png", sobelResult)

2、 运行结果如下

  1. 实验原图

  2. 实验结果图

五、 实验总结

  • 1、 掌握了编写含颜色算子图像处理、Susan与Harris角点图像检测、sobel边缘算子图像检测的程序编写方法。
  • 2、 通过实验、对于边缘检测算子与角点检测算子有了进一步的掌握。