Auto DR and Auto WB in Batch

[tomczak]

I'm experimenting with auto-corrections in batch. In Levels and Colour full range, Colour Balance Auto/Auto Full range, Auto Range and Equalize (and I'm not sure what else) the corrections depend on the image. Before I go too far, in Batch mode would those transformations get trained on the first image in the batch or should the corrections change based on every image being processed in turn? Thanks!

[jsachs]

They all change with the input image.

[tomczak]

Thank you. One thing I can't figure out is how Levels and Colour knows that the full range has been requested (or how I can know that it knows) as opposed to the image being already at full range (which means that the next one in the batch will not be stretched). This is from Image Info:

levels_and_color 0 colorspace 0 v1 0 v2 255 brightness 0 contrast 0 saturation 0 colorbal ffffff probesize 2 display 2 sync_scroll 1 auto_preview 1 amount 1 -1 0 0 ″Amount: ″ ″Mask″

[jsachs]

Sorry, I was incorrect earlier. Levels and Color Full Range is not re-applied when changing input images as this would require a somewhat different user interface -- you would have to use Autorange for to expand images to full range for changing input images. Autorange might do a little better job as Levels and Color uses fixed, arbitrary thresholds to compute its black and white points.

[tomczak]

Thanks again! I did an experiment, which confirms what Jonathan explained, but serves as a kind of a warning that if Levels and Colour is used in a batch, the dynamic range slider settings of the training image will stay constant when applied to all subsequent images.

So, for example, if the first image does not extend to full range, the dynamic sliders will automatically match the actual dynamic range ticks of this image and will stay there for the rest of the batch. That means that all subsequent images will have their white and black points moved according to those set limits, sometimes reducing, sometimes stretching their existing white and black points.

But if the dynamic range sliders are consciously moved away (or not moved) from the tick marks of the first image (including stretching it to full dynamic range either by adjusting the Dynamic Range sliders by hand, by Shift-clicking on it or by clicking the Full Dynamic Range in the transformation settings), all subsequent images will be stretched to those constant limits, chosen in the first/training image.

So to me, if Levels and Colour is to be used in a batch, the most sensible setting is to stretch the first/training image to 0-100% or such dynamic range, rather than relying on the chanced actual dynamic range of the first image, with the understanding that all subsequent images in the batch will be stretched to the same limits as well.

If stretching the dynamic range of ALL the images in the batch is not intended, I would stay away from Levels and Colour and use e.g. Adjust instead.