Re-compress dicom video

[RobinFrcd]

Hi,
I'm working on dicom video PixelData, and want to save it back to disk.

The thing is, while reading the data, I decompress it. Is there an easy way to compress it back before saving it to disk ? Because the new file is 25x bigger than the original file, for now.

import os
import pydicom

PATH = '....dcm'

dcm = pydicom.read_file(PATH)
print(os.path.getsize(PATH)/1e6)
# 2.466612
print(dcm.pixel_array.shape)
# (56, 480, 640, 3)


if dcm.file_meta.TransferSyntaxUID.is_compressed is True:
    dcm.decompress()
    
dcm.PixelData = dcm.pixel_array.tobytes()
dcm.PhotometricInterpretation  = "RGB"

dcm.save_as(PATH+'.copy')
print(os.path.getsize(PATH+'.copy')/1e6)
# 51.61107

I've tried to compress it back with dcm.compress but it raises ValueError: The Transfer Syntax UID element in 'FileDataset.file_meta' is compressed but the pixel data has been decompressed on save.

---

I've also tried to use PIL:

dcm.PixelData = encapsulate([Image.fromarray(f, 'I;16').tobytes() for f in cleaned_pixels])
dcm.compress(RLELossless)
dcm.file_meta.TransferSyntaxUID = RLELossless
dcm.is_decompressed = False

To file size still goes from 4MB to 80MB tho, and is not valid anymore: ValueError: The amount of decoded RLE segment data doesn't match the expected amount (321207 vs. 480000 bytes).

---

I tried

dcm.PhotometricInterpretation = 'RGB'
dcm.compress(RLELossless, cleaned_pixels)
dcm.is_decompressed = False

which worked, but I'm still going from 4MB to 107MB.

Thanks

[scaramallion]

pydicom doesn't natively support compression (except for RLE Lossless, which as you've seen isn't a very good compressor). You need to use third party libraries to perform the actual compression, then you can use pydicom to add the compressed data to your datasets.

[scaramallion]

@YaroslavPavlovich could you provide more details, please? Initial transfer syntax would be a good start, as well as what transfer syntax you would like to recompress to.