How can i generate a video from dicom frames with opencv?

[yuanzicheng]

I want to read the image frames from a dicom file, and use opencv to convert all the frames to a video.

My code is as follows:

from pydicom import dcmread
import cv2


def conver_dicom_to_video():
    dcm_file = "./test/2.dcm"
    
    output = dcm_file.replace(".dcm", ".mp4")
    fourcc = cv2.VideoWriter_fourcc(*'mp4v')
    fps = 2
    size = (512, 512)
    video = None
    
    dataset = dcmread(dcm_file)
    arr = dataset.pixel_array
    
    for i in range(10): # dataset.pixel_array.shape[0]
        print(i)
        img = arr[i]

        # save the frame image to disk first
        cv2.imwrite("{}.png".format(i), img)

        # read the image from disk
        img = cv2.imread("{}.png".format(i))

        if i == 0:
            # get the actual size of the generated images
            size = (img.shape[0], img.shape[1])
            video = cv2.VideoWriter(output, fourcc, fps, size)
            
        video.write(img)
    video.release()

When i save images first, and then use opencv to read the images and write them to a video, it works.

However, i think this is not a good way to save images to disk first. I tried to write the frames to the video directly, but i cloud not get the expected video file, the generated .mp4 file was always 1kb in size.

The code is as follows:

from pydicom import dcmread
import cv2


def conver_dicom_to_video():
    dcm_file = "./test/2.dcm"
    
    output = dcm_file.replace(".dcm", ".mp4")
    fourcc = cv2.VideoWriter_fourcc(*'mp4v')
    fps = 2
    size = (512, 512)
    video = cv2.VideoWriter(output, fourcc, fps, size)
    
    dataset = dcmread(dcm_file)
    arr = dataset.pixel_array
    
    for i in range(10): # dataset.pixel_array.shape[0]
        print(i)
        img = arr[i]
        video.write(img)
    video.release()

Cloud you please give me some help, and is there any way to get a video for the frames directly? Looking forward to your reply!

[darcymason]

I had a quick look at this to see if I could see any obvious issues. The only thing I wondered was whether the unaltered frame is a correct parameter to VideoWriter.write(). In trying to look that up, I saw one post that seems to succeed in doing what you are looking for, but to .avi file type. Perhaps that post can help.

[yuanzicheng]

Hmmm... That post is not useful for my case. Because it use OpenCV to 'split' RGB type image, but i have a look at the tag 'Photometric Interpretation', it's value is 'MONOCHROME2', so the functions cv2.split() and cv2.merge() are not apply to the frame which is not a RGB type image.

from pydicom import dcmread
import cv2
import numpy as np


def convert_dicom_to_video():
    dcm_file = "./test/2.dcm"
    
    output = dcm_file.replace(".dcm", ".avi")
    fourcc = cv2.VideoWriter_fourcc(*'MJPG')
    fps = 2
    size = (512, 512)
    video = cv2.VideoWriter(output, fourcc, fps, size)
    
    dataset = dcmread(dcm_file)
    print("Photometric Interpretation:", dataset.PhotometricInterpretation)
    
    for i in range(10): # dataset.pixel_array.shape[0]
        print(i)
        img = dataset.pixel_array[i]
        img = np.array(img)
        r, g, b = cv2.split(img)
        img = cv2.merge([b, g, r])
        img = cv2.resize(img, (512, 512))
        video.write(img)
    video.release()

Running result:

PS D:\code\dicom> python .\test.py
Photometric Interpretation: MONOCHROME2
0
Traceback (most recent call last):
  File "D:\code\dicom\test.py", line 8, in <module>
    convert_dicom_to_video()
  File "D:\code\dicom\test\dicom_to_video.py", line 22, in convert_dicom_to_video
    r, g, b = cv2.split(img)
ValueError: not enough values to unpack (expected 3, got 1)

Perhaps I should learn more about OpenCV...

[darcymason]

I don't have any experience with OpenCV, but I imagine there is either a flag to tell it you have greyscale data, or a way to convert grey to RGB. Hopefully you can find something with a little digging...

[yuanzicheng]

I got the solution! The parameter isColor should be set to 0 when calling the cv2.VideoWriter(), if the frames are greyscale images.

from pydicom import dcmread
import cv2


def convert_dicom_to_video():
    dcm_file = "./test/2.dcm"
    
    output = dcm_file.replace(".dcm", ".mp4")
    fourcc = cv2.VideoWriter_fourcc(*'mp4v')
    fps = 10
    size = (512, 512)
    video = cv2.VideoWriter(output, fourcc, fps, size, isColor=0)
    
    dataset = dcmread(dcm_file)
    for i in range(dataset.pixel_array.shape[0]):
        img = dataset.pixel_array[i]
        video.write(img)
    video.release()

Thanks a lot for your kindly help.

[darcymason]

Glad you solved it, and thanks for posting the code back here so others can find it.

[ray-lee-94]

@yuanzicheng Hello, How to get the duration and write the video in real FPS? I saw you set the FPS to 2.

[yuanzicheng]

The fps(Frames Per Second) of a video is decided by yourself.