Convert OpenEXR to JPEG using OpenImageIO(OIIO) in Python

Mastaka Koike — Sun, 27 Nov 2022 12:56:37 +0000

Self Introduction

I am Masataka @plinecom, a VFX pipeline engineer at digitalbigmo Inc.

What is this?

I'm an OpenEXR pusher, but I often need help to preview OpenEXR easily, so I wanted to convert it per JPEG. But I already have tens of millions of JPEGs, which will increase in the future, so I want to process them in Python. I discovered that OpenImageIO is a convenient way to do this, so I used OpenCV. And since it handles image data in NumPy arrays, it will be helpful later. It needed to support DPX as well.

Code (low-level API version)

import OpenImageIO as oiio
import glob

if __name__ == '__main__':

    for path in glob.glob('/foo/bar/*.exr'):
        print(path)
        jpg_path = path.replace('.exr', '.jpg')

        #Read File
        ImageInput.open(path)
        spec = inp.spec()
        pixels = inp.read_image("float")
        inp.close()

        #Write File
        ImageOutput.create(jpg_path)
        out.open(jpg_path, spec)
        out.write_image(pixels)
        out.close()

Code (high-level API version).

import OpenImageIO as oiio
import glob

if __name__ == '__main__':

    for path in glob.glob('/foo/bar/*.exr'):
        print(path)
        jpg_path = path.replace('.exr', '.jpg')

        #Read File
        buf = oiio.ImageBuf(path)

        #Write File
        buf.write(jpg_path)

Required Python external modules.

OpenImageIO
NumPy

I also prepared a Dockerfile.

FROM rockylinux:8
RUN dnf install -y which python3 epel-release
RUN dnf config-manager --set-enabled powertools
RUN dnf install -y python3-openimageio python3-numpy

docker pull plinecom/py_oiio

Description of the low-level API version of the code

It is faster to use the low-level API if you are trying to get a NumPy array. (There is also a way to take a NumPy array from the high-level API.)

Instructions on how to read EXR files.

        #Read File
        inp = oiio.ImageInput.open(path)
        spec = inp.spec()
        pixels = inp.read_image("float")
        inp.close()

The object pixels you can retrieve from the read_image() function is a NumPy array. Since this is OpeEXR, we instruct it to output the image data as a float type.

Instructions for exporting and converting image data

        #Write File
        out = oiio.ImageOutput.create(jpg_path)
        out.open(jpg_path, spec)
        out.write_image(pixels)
        out.close()

Specify the destination and file name to write to. The image format is automatically processed based on the file extension, so if you change the extension, OpenImageIO will output to a supported format.

Description of the high-level API version of the code

Nothing special. There is only one line for each read/write, and you should be able to figure it out if you see it.

Advertising

digitalbigmo Inc. sells skin beauty plug-ins and does video VFX production work. If you are interested, please visit our web page. Let's work together.

https://digitalbigmo.com

References

OpenImageIO main house (English)
OpenImageIO Reference (English)

Convert OpenEXR file compression method in Python

Mastaka Koike — Wed, 23 Nov 2022 08:54:05 +0000

Self Introduction

I am Masataka @plinecom, a pipeline engineer at digitalbigmo, Inc.

What is this?

I was given tens of thousands of large uncompressed OpenEXR files at work. Since OpenEXR has a lossless compression option, I want to convert this large number of uncompressed OpenEXR files to lossless compressed ZIP OpenEXR.
I will record how to batch-process them using Python.

For now, the code

import OpenEXR
import Imath
import glob
import os

if __name__ == '__main__':

    for path in glob.glob('/foo/bar/*.exr'):
        print(path)
        dir_name = os.path.dirname(path)
        base_name = os.path.basename(path)

        InputFile(path) # Read EXR file
        header = ret.header() # Get OpenEXR header info
        print(header)
        # Change Compression Type
        header['compression'] = Imath.Compression(Imath.Compression.ZIP_COMPRESSION)
        print(header)

        pix = {}
        for ch in header['channels']: # Get channel name from header
            pix[ch] = ret.channel(ch) # Get channel image data
        ret.close() # Close Read File handler

        # Open the write file and Set header info
        OutputFile(dir_name + '/' + base_name, header)
        exr.writePixels(pix) # Write Pixel Data
        exr.close() # Close Write File handler

Required Python external modules

OpenEXR

pip install OpenEXR

You will install the download from PyPI using pip, and the Imath module with it.
However, you will need to install Python's header file for the build process. If you cannot prepare it, it is easier to build it from Python.

I also prepared a Dockerfile. Please forgive me if some files are downloaded from rpmfusion, which is a bit weird. If you have a better way, please let me know.

FROM rockylinux:8
RUN dnf install -y which python3 python3-devel gcc gcc-c++ epel-release zlib-devel
RUN dnf config-manager --set-enabled powertools
RUN dnf localinstall -y --nogpgcheck https://download1.rpmfusion.org/free/el/rpmfusion-free-release-8.noarch.rpm
RUN dnf install -y OpenEXR-devel
RUN pip3 install OpenEXR
RUN dnf install -y --nogpgcheck RUN dnf install -y

docker pull plinecom/py_exr_zip

Explanation of code

Load EXR file

        InputFile(path) # Read EXR file

Specify the compression method

        header = ret.header()
        # Change Compression Type
        header['compression'] = Imath.Compression(Imath.Compression.ZIP_COMPRESSION)

If you write the compression method directly in the text and assign it, it will fail; you must pass an Imath.Compression object initialized with the Imath.Compression.* enumerator. In this case, we want to compress ZIP, so we pass Imath.Compression.ZIP_COMPRESSION. Since we do not change anything other than the compression method, we use the header information of the original file for the rest of the header information.

Writing files

        ## Open the write file and Set header info
        OutputFile(dir_name + '/' + base_name, header)
        exr.writePixels(pix) # Write Pixel Data
        exr.close() # Close Write File handler

Create a file based on the diverted header, specifying the destination to write to, write pixel data, and close the file.

Explanation of OpenEXR compression options

Lossless compression options.

RLE

Run-length compression. It is a simple compression process but very fast because it takes only one pass and requires very little computation. It is very fast because it is a simple compression process that takes only one pass and requires very little computation. Since this alone reduces the size of the file by about 90%, there is no reason to choose uncompressed. ~~~ so don't send me uncompressed OpenEXR files. ~~~

ZIP

You all know ZIP compression. It compresses an image by dividing it into 16 lines, and it can reduce the size to about half of the uncompressed size.

ZIPS

ZIP compresses images by 16 lines, but ZIPS compresses by 1 line. S is for scan lines, not plural S. I heard a rumor that ZIPS compresses images faster with compositing software such as Nuke, but I have yet to feel any difference. Compression efficiency is lower than ZIP.

PIZ

PIZ compresses image data in the frequency domain using a wavelet transform, which is upward compatible with the Fourier transform. While it offers the best compression ratio, it is also very computationally intensive, and the sacrifice in computation time is too great for a compression that is a few percent smaller than ZIP compression.

Compression that can be lossless or lossy.

PXR24

PXR24 compresses data with quantization reduced to 24 bits. 16-bit half-precision floating-point data and 16-bit integer data below 24 bits are lossless compressions, but 32-bit single-precision floating-point data are lossy compression. You should handle this point cautiously, so do not select this option if you are unsure.

Lossy compression option

Other lossy compression options are as follows.

B44
B44A
DWAA
DWAB

However, most of us use OpenEXR for high quality. We will not use lossy compression, which may cause an accident due to different decompression results at the delivery destination. If you are a large studio that produces your master's, it may be an option in terms of storage efficiency.

digitalbigmo Inc. sells beautiful skin plug-ins and provides video VFX production services. If you are interested, please visit our web page. Let's work together.

https://digitalbigmo.com

References

About Python's OpenEXR module (English)

OpenEXR main house (English)

Wikipedia's OpenEXR page

http://yamagishi-2bit.blogspot.com/2020/05/vfxexr.html

DEV Community: Mastaka Koike

Convert OpenEXR to JPEG using OpenImageIO(OIIO) in Python

Self Introduction

What is this?

Code (low-level API version)

Code (high-level API version).

Required Python external modules.

Description of the low-level API version of the code

Instructions on how to read EXR files.

Instructions for exporting and converting image data

Description of the high-level API version of the code

Advertising

References

Convert OpenEXR file compression method in Python

Self Introduction

What is this?

For now, the code

Required Python external modules

Explanation of code

Load EXR file

Specify the compression method

Writing files

Explanation of OpenEXR compression options

Lossless compression options.

RLE

ZIP

ZIPS

PIZ

Compression that can be lossless or lossy.

PXR24

Lossy compression option

advertisement

References