stripe artifacts with "write to disk"

[schmiedc]

Hi,

I am running the Grid/Collection stitching plugin on a dataset with 3 channels.
When running the plugin with "write to disk" I get stripe artifacts.

However when I run it with "Fuse and display" these stripes are not visible.

This is the macro I am running:

run("Grid/Collection stitching", "type=[Positions from file] " +
        "order=[Defined by image metadata] " +
        "multi_series_file=" + dir + IJ.pad(filedir, 4) + File.separator + IJ.pad(filedir, 4) + "-"  + file + ".czi " +
        "fusion_method=[Linear Blending] regression_threshold=0.35 max/avg_displacement_threshold=2.50 " +
        "absolute_displacement_threshold=3.50 compute_overlap increase_overlap=0 " +
        "subpixel_accuracy computation_parameters=[Save computation time (but use more RAM)] " +
        //"image_output=[Fuse and display]");
        "image_output=[Write to disk] output_directory=[" + fileoutput + "]");

Fiji and ImageJ are up to date.
I run it on ubuntu 16.04.

Cheers,
Christopher

[imagejan]

@schmiedc @ctrueden is this still the case?
People are using image_output=[Write to disk] in various examples, see e.g. https://github.com/uw-loci/automatic-histology-registration-pyimagej/blob/8ad405170ec46dccbdc1c20fbbeb6eaff47b8b76/pseudo_modality.ipynb and imagej/pyimagej#192.

If the outputs of image_output=[Fuse and display] and image_output=[Write to disk] differ, we should warn and recommend using the Fusion.fuse API directly where possible.

[ctrueden]

@imagejan I'm sorry, I don't know.

@binli123 Heads up: your workflows that use the Grid/Collection stitching's "Write to disk" feature might be writing out suboptimal data, compared to using "Fuse and display" in memory. Would it be easy for you to check this?

[binli123]

I checked some recent stitching results obtained using this plugin with the Write to disk option via PyImagJ. They do not seem to have this artifact.

Single channel multiphoton:

A brighter example:

RGB image:

The yellow box marks the boundary of a tile.
The same image from the first example stitched with the the Fuse to display option in ImageJ:

Overall the transition between boundaries looks quite smooth.

A snippet of the code I used:

    params = {'type': 'Positions from file', 'order': 'Defined by TileConfiguration', 
            'directory':stitch_folder, 'ayout_file': 'TileConfiguration.txt', 
            'fusion_method': 'Linear Blending', 'regression_threshold': '0.30', 
            'max/avg_displacement_threshold':'2.50', 'absolute_displacement_threshold': '3.50', 
            'compute_overlap':False, 'computation_parameters': 'Save computation time (but use more RAM)', 
            'image_output': 'Write to disk', 'output_directory': temp_channel_folder}
    plugin = "Grid/Collection stitching"
    ij.py.run_plugin(plugin, params)

[ctrueden]

Great, thanks @binli123! I will close this then. We can reopen if anyone has a recipe to reproduce with the latest version.