Lab Exercise 3
一. 简答题（共1题，100分）
1. (简答题, 100分)
Students are required to submit your solutions before 0:00am on October 31st. You should only upload two files (do not zip), your notebook ".ipynb" file and PDF file generated after notebook execution.
1. Convert an image to grayscale
Select an interesting RGB image, and convert the image to grayscale according to the following formula
Y = 0.2126*R + 0.7152*G + 0.0722*B, and plot the result.
2. Alpha Blending
Choose multiple photos as your wish, set one (or several) of them to the background, and the others to the foreground, and use the alpha channel to design the composition formula. For example, output = img1 * alpha1 + img2 * alpha2+.. .+imgN*alphaN, alpha1+alpha2+...+alphaN=1
1) Use your composition formula to generate a new image and display it
2) Explain what effect you expect to achieve by using your composition formula? , what are the characteristics? , what problem did you encounter? How did you solve it.
3. 4-connectivity and m-connectivity labeling
Given a binary image as input, the task is to label all white pixels (255) as connected components according to the 4-connectivity and m-connectivity rules. Different areas should be colored differently.
1) Students are required to implement the 4-connectivity and m-connectivity labeling algorithm (It is not allowed to directly or indirectly call third-party functions to complete 4-connectivity and m-connectivity labeling). The algorithm should support at least 100 labels. It takes any binary image as input and outputs the 4-connectivity and m-connectivity labelled images. The output should be similar to the images shown above.
2) Follows the requirements below, verify your algorithm and display the results:
a. Take the following image as input, generate the 4-connectivity and m-connectivity labelled images, and plot them into one row and three columns as the example shown above.
b. Use test_image=generate_image(4,16) to generate a test image, apply your algorithm on it to produce the 4-connectivity and m-connectivity labelled images, and plot them into one row and three columns as the example shown above.
c. Use test_image=generate_image(10,512) to generate a test image, do the same as the step b.
import time
import numpy as np
from skimage import filters
def generate_image(n,l):
print("n=",n," l=",l)
print(time.gmtime())
print(time.time())
timestamp = int(time.time())
print(timestamp)
np.random.seed(timestamp)
im = np.zeros((l, l))
points = l * np.random.random((2, n ** 2))
im[(points[0]).astype(np.int32), (points[1]).astype(np.int32)] = 1
im = filters.gaussian(im, sigma= l / (5. * n))
im = im > 0.7 * im.mean()
return im
3）As shown in the figure below, there is a hole in the region. Please elaborate how to detect a hole in such region? (Students who are capable can try to implement your solution.)
作答（别不完全对，进展示我的）
1. Convert an image to grayscale
Select an interesting RGB image, and convert the image to grayscale according to the following formula
Y = 0.2126R + 0.7152G + 0.0722*B, and plot the result.
import CV2 as cv
import numpy as np
import matplotlib.pyplot as plt
from PIL import Image
import matplotlib.image as image
import os
#展示图片函数
def show(img):
if img.ndim==2:
plt.imshow(img,cmap='gray')
else:
plt.imshow(cv.cvtColor(img,cv.COLOR_BGR2RGB))
test=cv.imread("Image/date_suqing.png")
show(test)
#使用公式转换图片 首先分离通道
b,g,r = cv.split(test)
img=b*0.0722+g*0.7152+r*0.2126
show(img)
print(test[0])
print(b)
2. Alpha Blending
Choose multiple photos as your wish, set one (or several) of them to the background, and the others to the foreground, and use the alpha channel to design the composition formula. For example, output = img1 * alpha1 + img2 * alpha2+.. .+imgN*alphaN, alpha1+alpha2+...+alphaN=1
1) Use your composition formula to generate a new image and display it
2) Explain what effect you expect to achieve by using your composition formula? , what are the characteristics? , what problem did you encounter? How did you solve it.
#大小规定
def resize_img(img,width,high):
re=cv.resize(img,(int(width),int(high)))
return re
#选择3张图片
img1=cv.imread("Image/date_suqing.png")
img2=cv.imread("Image/img2.png")
img3=cv.imread("Image/img3.png")
x,y=img1.shape[0:2]
img2=resize_img(img2,y,x)
img3=resize_img(img3,y,x)
images = [img1,img2,img3]
imgs=np.hstack([img1,img2,img3])
show(imgs)
foreground = img2.astype(float)
foreground2 = img3.astype(float)
background0 = img1.astype(float)
alpha1 = img2.astype(float)/(255*2)
alpha2 = img3.astype(float)/(255*2)
foreground = cv.multiply(alpha1, foreground)
foreground2 = cv.multiply(alpha2, foreground2)
background1 = cv.multiply(1.0 - alpha1, background0)
background2 = cv.multiply(1.0 - alpha2, background0)
outImage = cv.add(foreground, background1)
# cv.imshow("outImg", outImage/255)
# outImage1 = cv.add(foreground2, background2)
# cv.imshow("outImg1", outImage1/255)
outImage2 = cv.add(foreground2,outImage)
cv.imwrite("outimage.jpg",outImage2)
show(cv.imread('outimage.jpg'))
explain :
期望就是背景放在最后一层，前景在背景基础上能继承，都可以得到较合理的展示，使用相乘的结果，前景取了两张图片，设置两个alpha，来看图片，特点就是，背景图比前景更明显，两张前景图显得隐隐约约，遇到的问题是这样写到底算对不对呢
3. 4-connectivity and m-connectivity labeling
# 4-connectivity
image=cv.imread('Image/464c345f241b858ea278139190664b9.jpg')
show(image)
# 4-connectivity labeling
def four_cc_label(img):
# print(len(img.shape))
if(len(img.shape)==3):
height, width, channel = img.shape
else:#查看通道数
# im 为单通道图像 image为生成的三通道图像
img = img[:, :, np.newaxis]
img = img.repeat([3], axis=2)
height, width, channel = img.shape
label = np.zeros((height, width), dtype=np.int32)
LUT = np.zeros(height * width,dtype=np.uint8)
# 色彩列表 支持100个？ ，不如for循环他
COLORS = np.array([[0, 0, 255], [0, 255, 0], [255, 0, 0],
[255, 255, 0], [255, 0, 255], [0, 255, 255],
[125, 0, 255], [0, 255, 125], [255, 0, 125],
[255, 255, 125], [255, 125, 255], [0, 125, 255]])
for ila in range(256):
s=np.array([[255,0,ila],[125,0,ila]])
COLORS=np.append(COLORS,s,axis=0)
COLORS.astype(np.uint8)
#应该这样支持260左右
out = np.zeros((height, width, channel), dtype=np.uint8)
label[img[:,:, 0] > 0] = 1
n = 1
for y in range(height):
for x in range(width):
if label[y, x] == 0:
continue
c3 = label[max(y - 1, 0), x]
c5 = label[y, max(x - 1, 0)]
if c3 < 2 and c5 < 2:
n += 1
label[y, x] = n
else:
_vs = [c3, c5]
vs = [a for a in _vs if a > 1]
v = min(vs)
label[y, x] = v
minv = v
for _v in vs:
if LUT[_v] != 0:
minv = min(minv, LUT[_v])
for _v in vs:
LUT[_v] = minv
count = 1
for l in range(2, n + 1):
flag = True
for i in range(n + 1):
if LUT[i] == l:
if flag:
count += 1
flag = False
LUT[i] = count
# print(len(LUT[2:]))
for index, lut in enumerate(LUT[2:]):
# print(lut)
out[label == (index + 2)] = COLORS[lut - 2]
return out
#m-connectivity labeling
#find father and update
import CV2
import numpy as np
def four_cc_label2(img):
if(len(img.shape)==3):
height, width, channel = img.shape
else:#查看通道数
# im 为单通道图像 image为生成的三通道图像
img = img[:, :, np.newaxis]
img = img.repeat([3], axis=2)
height, width, channel = img.shape
label = np.zeros((height, width), dtype=np.int32)
LUT = np.zeros(height * width,dtype=np.uint8)
label = np.zeros((height, width), dtype=np.int32)
LUT = np.zeros(height * width,dtype=np.uint8)
COLORS = np.array([[0, 0, 255], [0, 255, 0], [255, 0, 0],
[255, 255, 0], [255, 0, 255], [0, 255, 255]],dtype=np.uint8)
out = np.zeros((height, width, channel), dtype=np.uint8)
label[img[:,:, 0] > 0] = 1
n = 1
for y in range(height):
for x in range(width):
if label[y, x] == 0:
continue
c2 = label[max(y - 1, 0), min(x + 1, width - 1)]
c3 = label[max(y - 1, 0), x]
c4 = label[max(y - 1, 0), max(x - 1, 0)]
c5 = label[y, max(x - 1, 0)]
if c3 < 2 and c5 < 2 and c2 < 2 and c4 < 2:
n += 1
label[y, x] = n
else:
_vs = [c3, c5, c2, c4]
vs = [a for a in _vs if a > 1]
v = min(vs)
label[y, x] = v
minv = v
for _v in vs:
if LUT[_v] != 0:
minv = min(minv, LUT[_v])
for _v in vs:
LUT[_v] = minv
count = 1
for l in range(2, n + 1):
flag = True
for i in range(n + 1):
if LUT[i] == l:
if flag:
count += 1
flag = False
LUT[i] = count
for i, lut in enumerate(LUT[2:]):
out[label == (i + 2)] = COLORS[lut - 2]
return out
prignal=cv.imread('image/bd33bf3400112b949021bfd582ce11df.png')
out = four_cc_label(prignal)
cv.imwrite("outimage2.jpg",out)
out2=cv.imread('outimage2.jpg')
out5 = four_cc_label2(prignal)
cv.imwrite("outimage5.jpg",out)
out5=cv.imread('outimage5.jpg')
# show(np.hstack([prignal,out2]))
plt.figure()#创建画布
plt.subplot(1,3,1)
plt.title("binary input")
plt.imshow(prignal)
plt.subplot(1,3,2)
plt.title("4-connectivity labeling")
plt.imshow(out2)
plt.subplot(1,3,3)
plt.title("m-connectivity labeling")
plt.imshow(out5)
plt.show()
import time
import numpy as np
from skimage import filters
def generate_image(n,l):
print("n=",n," l=",l)
print(time.gmtime())
print(time.time())
timestamp = int(time.time())
print(timestamp)
np.random.seed(timestamp)
im = np.zeros((l, l))
points = l * np.random.random((2, n ** 2))
im[(points[0]).astype(np.int32), (points[1]).astype(np.int32)] = 1
im = filters.gaussian(im, sigma= l / (5. * n))
im = im > 0.7 * im.mean()
return im
test=generate_image(4,16)
#将二维图像变为三维但不影响其图像显示
out = four_cc_label(test)
cv.imwrite("outimage3.jpg",out)
out3=cv.imread('outimage3.jpg')
out = four_cc_label2(test)
cv.imwrite("outimage6.jpg",out)
out6=cv.imread('outimage6.jpg')
plt.figure()#创建画布
plt.subplot(1,3,1)
plt.title("binary input")
plt.imshow(test)
plt.subplot(1,3,2)
plt.title("4-connectivity labeling")
plt.imshow(out3)
plt.subplot(1,3,3)
plt.title("m-connectivity labeling")
plt.imshow(out6)
plt.show()
#c.Use test_image=generate_image(10,512) to generate a test image, do the same as the step b.
test=generate_image(10,512)
#将二维图像变为三维但不影响其图像显示
out = four_cc_label(test)
cv.imwrite("outimage4.jpg",out)
out4=cv.imread('outimage4.jpg')
plt.figure()#创建画布
plt.subplot(1,2,1)
plt.title("binary input")
plt.imshow(test)
plt.subplot(1,2,2)
plt.title("4-connectivity labeling")
plt.imshow(out4)
plt.show()
#3）As shown in the figure below, there is a hole in the region. Please elaborate how to detect a hole in such region? (Students who are capable can try to implement your solution.)
#使用4领域标注法检测出一段差异的位置，即可标注出图像空白的部分
