OpenCV

Filtering

Averaging

blur = cv2.blur(img,(5,5))

Median Filtering

median = cv2.medianBlur(img,5)

Gaussian Filtering

blur = cv2.GaussianBlur(img,(5,5),0)

Bilateral Filtering

blur = cv2.bilateralFilter(img,9,75,75)

Merge Two Frames

#img == imread()
red_img = np.full((1080,1920,3), (0,0,255), np.uint8)
img_copy = cv2.addWeighted(src1=img, alpha=0.8, src2=red_img, beta=0.2, gamma=0)

Arithmetic Operations on Images

add:

cv2.add(img1, img2)

subtract:

 cv2.subtract(src1, src2)

Bitwise Operation

AND:

cv2.bitwise_and(img1, img2)

OR

cv2.bitwise_or(img1, img2)

XOR

cv2.bitwise_xor(img1, img2)

NOT

cv2.bitwise_not(img1)

Binary image:

src = cv2.imread("sample.jpg")
gray = cv2.cvtColor(src, cv2.COLOR_BGR2GRAY)
_, binary = cv2.threshold(gray, 127, 255, cv2.THRESH_BINARY)
and_img = cv2.bitwise_and(gray, binary)

Find object's minimal Rectangle

Get minimal rectangle:

rotate_rect = cv2.minAreaRect(array(obj_pts))
w = rotate_rect[1][0]
h = rotate_rect[1][1]
obj_center = rotate_rect[0]
angle = rotate_rect[2]

Get minimal rectangle points:

rect_pts = cv2.boxPoints(rotate_rect)

All about Contour

Post-Processing Contour

x, y, w, h = cv2.boundingRect(contour)

Get minimum rectangle box(rotated box)

rotateRect = cv2.minAreaRect(contour)
pts = cv2.boxPoints(rotateRect)  # rect 4-points

Get minimum circle

center, radius = cv2.minEnclosingCircle(contour)

Get minimum triangle

area, triangle = cv2.minEnclosingTriangle(points)

Get minimum ellipse

ellipse = cv2.fitEllipse(points)

Get line which crosses object center

line = cv2.fitLine(points, distType, param, reps, aeps, line)

reference1 [reference2]https://hoony-gunputer.tistory.com/entry/OpenCV-python-Contour [reference3]https://076923.github.io/posts/Python-opencv-21/

countNonZero

obj_area = cv2.countNonZero(binary_img)

Draw Contour

custom_roi_points are user defined points, and we use cv2.drawContours to draw its contour.

import numpy as np
import cv2
from numpy import array

img = np.zeros((512,512,3), np.uint8)
custom_roi_points = array([(50,50),(50,100),(100,100),(100,50)])

cv2.drawContours(img, [custom_roi_points], 0, (0,255,0), 2)
cv2.imshow('draw rect', img)
cv2.waitKey(0)
cv2.destroyAllWindows()

Result:

DrawRect

Draw polygon line

img = cv2.polylines(img, [pts], False, blue, 2) # open type poylgon
img = cv2.polylines(img, [pts], True, blue, 2) # close type poylgon

Draw Convex polygon

img = cv2.fillConvexPoly(img, [pts], blue)

Distance between polygon and points

If you want to check a point inside or outside of a polygon, use cv2.pointPolygonTest:

sample code:

import cv2
import numpy as np

blank_image = np.zeros((720,1280,3), np.uint8)
custom_roi_points= np.array([(430,360),(860,360),(860,710),(430,710)]) 

dist = cv2.pointPolygonTest(custom_roi_points, (500,600),True)
cv2.circle(blank_image,(500,600),1,[255,0,0],7)
cv2.drawContours(blank_image,[custom_roi_points], 0, (0,255,255),2)

print("dist= " + str(dist))
cv2.imshow('img',blank_image)
cv2.waitKey(0)
cv2.destroyAllWindows()

inshow setting

Scalable Window Size

cv2.namedWindow("result", cv2.WINDOW_NORMAL)