从 OpenCV-python 中的视频中保存感兴趣的帧

Question

The code below stores all frames from a video and then selects and stores frames of interest which in this case is a frame per 90 frames.

import cv2
import numpy as np
import matplotlib.pyplot as plt 


vidcap = cv2.VideoCapture("video .wmv")

success,image = vidcap.read()
count = 0
success = True
frames = []
while success:
    frames.append(image)
    success,image = vidcap.read()
    count += 1
 
vidcap.release()

#print(len(frames))

interval = 90

selectframes = []
framenum = []  #frame number of the selected frame 

for i in range(0,len(frames),interval):
    selectframes.append(frames[i])
    framenum.append(i)

While the code works as intended, I just wanted to know if there is a better way of doing it.

Answer 1

at the very least, you can do if count % 90 == 0: frames.append(image)

there is potential for optimization. OpenCV's VideoCapture has not just a read() method but also the pair of grab() and retrieve() methods, which is precisely what read() does. grab() only makes sure that the frame was acquired, retrieve() actually decodes it. this may or may not work on video files (depends on how well the backend/apiPreference implements it). calling grab() without retrieve() would save the program the effort of actually decoding each frame.

assuming grab() does less work than read(), you can try vidcap.grab() 90 times, once vidcap.retrieve(), and repeat.

Answer 2

You can try this one

import cv2
import os
count=1

vidcap = cv2.VideoCapture('video.wmv')
def getFrame(sec):
    vidcap.set(cv2.CAP_PROP_POS_MSEC,sec*1000)
    hasFrames,image = vidcap.read()
    if hasFrames:
        dim = (512, 512)
        resized = cv2.resize(image, dim, interpolation = cv2.INTER_AREA)
        cv2.imwrite("images/"+str(count)+".png", resized)
    return hasFrames
sec = 0
frameRate = 90/1000

success = getFrame(sec)
while success:
    count = count + 1
    sec = sec + frameRate
    sec = round(sec, 2)
    success = getFrame(sec)