Python: Create video out of processed images

Question

i have a video where the front view of a car was recorded. The file is an .mp4 and i want to process the single images so i can extract more information (Objects, Lane Lines etc.). The problem is, when i want to create a video out of the processed files, i get an error. Here is what i have done so far:

Opened the video with cv2.VideoCapture() - Works fine
Saved the single frames of the video with cv2.imwrite() - Works fine
Creating a video out of single frames with cv2.VideoWriter() - Works fine
Postprocessing the video with cv2.cvtColor(), cv2.GaussianBlur() and cv2.Canny() - Works fine
Creating a video out of the processed images - Does not work.

Here is the code i used:

enter code here
def process_image(image):
    gray = functions.grayscale(image)
    blur_gray = functions.gaussian_blur(gray, 5)
    canny_blur = functions.canny(blur_gray, 100, 200)

    return canny_blur

process_on =0
count =0

video= cv2.VideoWriter("output.avi", cv2.VideoWriter_fourcc(*"MJPG"), 10, (1600, 1200))

vidcap = cv2.VideoCapture('input.mp4')
success,image = vidcap.read()

success = True
while success:
  processed = process_image(image)
  video.write(processed)

This is the error i get:

OpenCV Error: Assertion failed (img.cols == width && img.rows == height*3) in cv::mjpeg::MotionJpegWriter::write, file D:\Build\OpenCV\opencv-3.2.0\modules\videoio\src\cap_mjpeg_encoder.cpp, line 834 Traceback (most recent call last): File "W:/Roborace/03_Information/10_Data/01_Montreal/camera/right_front_camera/01_Processed/Roborace_CAMERA_create_video.py", line 30, in video.write(processed) cv2.error: D:\Build\OpenCV\opencv-3.2.0\modules\videoio\src\cap_mjpeg_encoder.cpp:834: error: (-215) img.cols == width && img.rows == height*3 in function cv::mjpeg::MotionJpegWriter::write

My suggestion is: The normal images have 3 dimensions because of the RGB-color field. The processed images are only having one dimension. How can i adjust this in the cv2.VideoWriter function.

Thanks for your help

On StackOverflow, it's customary to upvote answers you find helpful (that's the little upwards pointing triangle next to the answer). Also, if it's an answer to your question and you deem it the best one, there's a little checkmark that let's you "accept" it. — anms_pro

alkasm alkasm · Accepted Answer · 2017-10-02T13:13:50

The VideoWriter() class only writes color images, not grayscale images unless you are on Windows, which it looks like you might be judging from the paths in your output. On Windows, you can pass the optional argument isColor=0 or isColor=False to the VideoWriter() constructor to write single-channel images. Otherwise, the simple solution is to just stack your grayscale frames into a three-channel image (you can use cv2.merge([gray, gray, gray]) and write that.

From the VideoWriter() docs:

Parameters:

isColor – If it is not zero, the encoder will expect and encode color frames, otherwise it will work with grayscale frames (the flag is currently supported on Windows only).

So by default, isColor=True and the flag cannot be changed on a non-Windows system. So simply doing:

video.write(cv2.merge([processed, processed, processed])

should patch you up. Even though the Windows variant allows writing grayscale frames, it may be better to use this second method for platform independence.

Also as Zindarod mentions in the comments below there are a number of other possible issues with your code here. I'm assuming you've pasted modified code that you weren't actually running here, or that you would have modified otherwise...if that's the case, please only post minimal, complete, and verifiable code examples.

First and foremost, your loop has no end condition, so it's indefinite. Secondly, you're hard-coding the frame size but VideoWriter() does not simply resize the images to that size. You must provide the size of the frame that you will pass into the VideoWriter(). Either resize the frame to the same size before writing to be sure, or create your VideoWriter using the frame size as defined in your VideoCapture() device (using the .get() methods for the frame size properties).

Additionally, you're reading only the first frame outside the loop. Maybe that was intentional, but if you want to process each frame of the video, you'll need to of course read them in a loop, process them, and then write them.

Lastly you should have some better error catching in your code. For e.g., see the OpenCV "Getting Started with Video" Python tutorial. The "Saving a Video" section has the proper checks and balances: run the loop while the video capture device is opened, and process and write the frame only if the frame was read properly; then once it is out of frames, the .read() method will return False, which will allow you to break out of the loop and then close the capture and writer devices. Note the ordering here---the VideoCapture() device will still be "opened" even when you've read the last frame, so you need to close out of the loop by checking the contents of the frame.

Python: Create video out of processed images

2 Answers