New batch processing options for image export

Image export options, and a corresponding image export dialog, have been implemented in PhotoFlow quite some time ago:

However, until now those options were not yet implemented in the batch processor. While introducing the image export options, I also had to partly change the way command-line parameters are passed to the batch processor.

As usual, up-to-date packages for Linux, MacOS and Windows 64 bits are available from here.



You can get an exhaustive help for the new batch processing options by typing

photoflow --batch --help

which will give you the following output:

$ photoflow --batch --help

Usage: photoflow --batch [--config=config_file.pfi] --export-opt=[export options] in_file out_file

Export options (comma-separated list):

  jpeg_quality=[0..100] (default: 80)

  jpeg_chroma_subsampling=0/1 (default: 0)

  jpeg_quant_table=default:medium:best (default: best)

  tiff_depth=8:16:32 (default: 16) tiff_depth=32 generates an image in floating-point format

  tiff_compress=0/1 (default: 0)

  width=X (width of the exported image in pixels)

  height=X (height of the exported image in pixels)

  sharpen_enabled=0/1 (default: 0) enable/disable post-resize sharpening step

  sharpen_radius=X (radius for the post-resize sharpening step)

  profile_type=X (default: sRGB) ICC profile for the exported image

    allowed values are:

      no_change: keep the image in the same colorspace used for processing (by default linear Rec2020)

      sRGB, Rec2020, AdobeRGB, ProPhoto, ACEScg, ACES: convert the image to the specified colorspace

      from_disk: use a custom ICC profile from disk

  profile_name="X.icc" path to profile from disk when "profile_type" is set to "from_disk"

  trc_type=standard:linear:perceptual:sRGB (default: standard) TRC of the output ICC profile

    the meaning of the values is:

      standard: use the standard TRC for the selected colorspace

        for example: sRGB TRC for the sRGB colorspace, gamma=2.2 for AdobeRGB, etc...

      linear: linear (gamma=1.0) TRC

      perceptual: same TRC as defined in the CIELab L channel specifications

      sRGB: same TRC as defined in the sRGB specifications

    Notice that this option has no effect when using a custom profile from disk

  intent=relative_colorimetric:perceptual:saturation:absolute_colorimetric (default: relative_colorimetric)

    Rendering intent for the output ICC conversion.

    Note that only relative and absolute colorimetric intents are implemented for the built-in colorspaces.

    Perceptual and saturation intents might be available when using LUT profiles from disk.

  bpc=0/1 (default: 1) enable/disable black point compensation in the output ICC conversion

Example:

  --export-opt=tiff_depth=16,tiff_compress=1,width=800,height=600,sharpen_enabled=1,sharpen_radius=0.5,\

    profile_type=Rec2020,trc_type=linear

As you can see, the export options are appended as a comma-separated list of parameter=value pairs. In addition, one or more .pfi configuration files can be specified by appending several --config options. The processing parameters in the .pfi files will be sequentially applied to the input image (the parameters of the first --config option are applied first), before exporting to the output format.

The last two command-line arguments are the names of the input and output image files. The format of the output image is automatically guessed based on the file extension (.tif or .tiff for TIFF files, .jpg or .jpeg for Jpeg files). Both uppercase and lowercase extensions are accepted.

The output file name supports a special syntax that simplifies the creation of output files with the same base-name of the input, but a different extension and/or destination path. For this, the special string %name% in the output file name gets replaced with the base name (without extension) of the input file. For example, the command

photoflow --batch Pictures/image.tif Pictures/Web/%name%.jpg

will convert the input image to Jpeg and save it as image.jpg in the Pictures/Web folder.

A note about image resizing

When resizing the image with the width and height export options, PhotoFlow uses the default vips_resize() function of the underlying VIPS library, with a lanczos3 kernel. This function has been tuned to provide a very good balance between sharpness of the resized image and presence of moire artefacts, as explained here and here. Moreover, the input image is always promoted to 32bit floating-point format and converted to linear Rec.2020 colorspace upon loading. Hence, and unless you do some colorspace manipulations via additional .pfi files passed through --config options, all resizing operations will be performed in linear gamma encoding for optimal results. The image data is converted to the output colorspace only after the resizing step.