visualmic.py

import argparse
import dtcwt
from scipy import ndimage
from scipy import signal
import numpy as np
import cv2
import cmath
from scipy.io.wavfile import write

def maxTime(a,b):
		c = np.zeros_like(a)
		length = a.size
		for shift in range(length):
				c[shift]=np.dot(a,np.roll(b,shift))
		return np.argmax(c)

def npTowav(np_file, output_name):
		# Samples per second(sampling frequency of the audio file)
		sps = 1
		waveform_integers = np.int16(np_file * 32767)

		# Write the .wav file
		write(output_name, sps, waveform_integers)
		print(f"Output saved to {output_name}")

def soundfromvid(input_data,frameCount,nlevels,orient,ref_no,ref_orient,ref_level):

	tr = dtcwt.Transform2d()
	ref_frame= tr.forward(input_data[ref_no],nlevels =nlevels)
	data = np.zeros((frameCount,nlevels,orient))

	for fc in range(frameCount):  
		frame = tr.forward(input_data[fc],nlevels =nlevels)
		for level in range(nlevels):
	
			xmax=len(frame.highpasses[level])
			ymax=len(frame.highpasses[level][0])
	
			for angle in range(orient):
	
				for x in range(xmax):
	
					for y in range(ymax):
	
						amp,phase=cmath.polar(frame.highpasses[level][x][y][angle])
						ref_phase=cmath.polar(ref_frame.highpasses[level][x][y][angle])[1]
	
						data[fc,level,angle]+=(amp*amp)*(phase-ref_phase)
	print("Transform complete")
	shift_matrix=np.zeros((nlevels,orient))
	ref_vector=data[:,ref_level,ref_orient].reshape(-1)
	for i in range(nlevels):
		for j in range(orient):
			shift_matrix[i,j]=maxTime(ref_vector,data[:,i,j].reshape(-1))
	
	sound_raw=np.zeros(frameCount)
	for fc in range(frameCount):
		for i in range(nlevels):
			for j in range(orient):
				sound_raw[fc]+=data[fc-int(shift_matrix[i,j]),i,j]
	p_min = sound_raw[0]
	p_max = sound_raw[0]
	for i in range(1,frameCount):
		if sound_raw[i]<p_min:
			p_min=sound_raw[i]
		if sound_raw[i]>p_max:
			p_max=sound_raw[i]
	sound_data=((2*sound_raw)-(p_min+p_max))/(p_max-p_min)
	
	return (sound_data)


def main():
	# Create the argument parser
	parser = argparse.ArgumentParser(description='Motion Magnification using 2D DTCWT')

	# Add arguments
	parser.add_argument('-i', '--input', required=True, help='Specify input video path')

	# Parse the command-line arguments
	args = parser.parse_args()
	filename = args.input

	cap = cv2.VideoCapture(filename)

	fps = cap.get(cv2.CAP_PROP_FPS)
	frameCount = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
	frameWidth = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
	frameHeight = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))

	print(f"frameCount, frameWidth, frameHeight")
	print(frameCount,frameWidth,frameHeight,sep="             ")

	nlevels=3
	orient=6
	ref_no=0
	ref_level=0
	ref_orient=0

	input_data = []


	fc = 0
	ret = True

	for fc in range(frameCount):

		ret, frame = cap.read()
		gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
		input_data.append(gray)

	cap.release()
	print("Video Loaded")


	npTowav(soundfromvid(input_data,frameCount,nlevels,orient,ref_no,ref_orient,ref_level), "sound.wav")


if __name__ == "__main__":
		main()