What Are The Basic Image Processing & Connected Component Analysis

 Basic Image Processing & Connected Component Analysis

Objective: 

The objective of this whole topic  is to introduce you with some transformation especially with respect to image processing and perform connected component labelling in images and to get an understanding of intensity level resolution. 

Description:

1) Rotation:

2) Connected Component Analysis:

Connected Component Analysis Connected Component Analysis or Labelling enables us to detect different objects from a binary image. Once different objects have been detected, we can perform a number of operations on them: from counting the number of total objects to counting the number of objects that are similar, from finding out the biggest object of the bunch to finding out the smallest and from finding out the closest pair of objects to finding out the farthest etc. Connected Component labelling procedure is as follows: Process the image from left to right, top to bottom: If the next pixel to process is If only one from top or left is 1, copy its label. then If both top and left are one and have the same label, copy it. If top and left they have different labels then Copy the smaller label now Update the equivalence table .Otherwise, assign a new label. simply Re-label with the smallest of equivalent labels.

Now there some task you have to perform let's start:

Task No 01:

1:Create a generic code that create a border around any landscape image as shown below. The length of right and left borders must be 10% of the original horizontal length of the image. The length of upper and lower border must be such that the image now have same number of rows as columns. Save the image. 

2:Read any image that you want and save it in gray scale. Now rotate the image that you have read. Write the image to the disk. Hint: you can use built in functions of opencv 

 3: For the image given below (provided with the lab handout), apply the connectected component labelling and count the total number of white objects. First threshold the images and then do connected component analysis

 

 

 

 

 

 

Task 01

Code

import cv2 as cv
borderType = cv.BORDER_CONSTANT
img= cv.imread('Picture5.png')              #read input image
vertical_length = img.shape[0]              #number of rows
horizontal_length = img.shape[1]            #number of columns
left = int(0.10 * horizontal_length)        #10 percent of horizontal length
top = 2*left
bottom = top
right = left

#add broder to image
new_img = cv.copyMakeBorder(img, top, bottom, left, right, borderType, None)

#save image with border image

cv.imwrite('image\img_with_Border_1.jpg',new_img)      

Output

 

 

 

 

 

 

Task 02

(a)

Code

import cv2 as cv
img = cv.imread('pic.png')                           #read original image                                       
gray_img = cv.cvtColor(img,cv.COLOR_BGR2GRAY)        #convert image to grayscale
cv.imwrite('image\img_in_gray_1.jpg',gray_img)       #save gray scaled image

Output

 

  

(b)

Code

import cv2 as cv
img = cv.imread('pic.png')                             #read original image

#rotate image by 90 degrees clockwise
img_rotate_90_clockwise = cv.rotate(img, cv.ROTATE_90_CLOCKWISE) 

cv.imwrite('image\Rotate_90_1.jpg',img_rotate_90_clockwise)   #save rotated image

Output

 THINK!! 

1. What will be the number of dimension of a grayscale image if opened as colored? 

2. Is image a list, tuple or an array? 

3. A black and white image can only have what gray levels?

 4. Is it possible to count objects which are inside an object using connected component analysis? 

5. Can we apply connected component analysis any image without doing any preprocessing