I tried to make the political broadcast video like IPPON Grand Prix (OpenCV: Python version)

Trigger

When I tried to binarize the characters in the video, only the person and the telop were extracted probably because the background was not in the natural environment and nothing was reflected, but if the background color was yellow I thought it would look like an IPPON Grand Prix, so I implemented it.

development of

import cv2
import numpy as np

if __name__ == '__main__':

	cap = cv2.VideoCapture('one_minutes.mp4')

	cap_width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
	cap_height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
	fps = cap.get(cv2.CAP_PROP_FPS)

	telop_height = 50

	fourcc = cv2.VideoWriter_fourcc('m','p','4','v')
	writer = cv2.VideoWriter('threshold_raw.mp4',fourcc, fps, (cap_width, cap_height + telop_height))

	kernel = np.ones((3,3),np.uint8)

	count = 0
	try :
		while True:
			if not cap.isOpened():
				break

			if cv2.waitKey(1) & 0xFF == ord('q'):
				break

			ret, frame = cap.read()

			if frame is None:
				break

			frame = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
			ret2, frame = cv2.threshold(frame, 0, 255, cv2.THRESH_OTSU)

			#Shrink->expansion
			frame = cv2.morphologyEx(frame, cv2.MORPH_OPEN, kernel)

			background = np.zeros((cap_height, cap_width, 3), np.uint8)
			background[:] = tuple((80,235,247))

			telop = np.zeros((telop_height, cap_width, 3), np.uint8)
			telop[:] = tuple((128,128,128))

			#Once monochrome(dim=1)From color(dim=3)Conversion to
			frame = cv2.cvtColor(frame, cv2.COLOR_GRAY2BGR)
			#Background color and composition
			frame = cv2.bitwise_and(frame, background)

			images = [frame, telop]

			frame = np.concatenate(images, axis=0)
			font = cv2.FONT_HERSHEY_SIMPLEX
			cv2.putText(frame, "{:.4f} [sec]".format(round(count/fps, 4)), 
						(cap_width - 250, cap_height + telop_height - 10), 
						font, 
						1, 
						(0, 0, 255), 
						2, 
						cv2.LINE_AA)

			writer.write(frame)
			count += 1

	except cv2.error as e:
		print(e)	

	writer.release()
	cap.release()

Supplement

ret2, frame = cv2.threshold(frame, 0, 255, cv2.THRESH_OTSU) It is binarized with. Here, Otsu's algorithm is used.

frame = cv2.morphologyEx(frame, cv2.MORPH_OPEN, kernel) Now, from the binarized image, shrink-> expand and the white spots are filled with black at the stage of contraction to remove noise.

background = np.zeros((cap_height, cap_width, 3), np.uint8)
			background[:] = tuple((80,235,247))
background[:] = tuple((80,235,247))

#Once monochrome(dim=1)From color(dim=3)Conversion to
frame = cv2.cvtColor(frame, cv2.COLOR_GRAY2BGR)
#Background color and composition
frame = cv2.bitwise_and(frame, background)

Creates a single yellow background color and composites it with the binarized image with bitwise_and.

result

No opening process (no processing)

With opening process

It may be a little confusing, but if you look at the names of the parties above, you can see that the noise disappears when the contraction is reduced.

in conclusion

I was able to separate the objects fairly cleanly, so next time I will try to extract the characters.

Link I tried to refer to

-opencv-python opening (Opening) -Opencv-python Bit unit processing -IPPON Grand Prix