Layout Transform

[Marpel]

I just put together a nine panel square image (using 3 source images across and 3 source images high), using the Layout Transform. Each individual source image is 2333 x 2333 and I set up the transform so each panel has a 100 pixel border (so the result is a 100 pixel outer border, then 100 pixel space between each panel image, both horizontally and vertically).

Everything worked out fine and the resultant image is 7399 pixels square (4 x 100 pixels plus 3 x 2333 pixels).

My question, though, is when I pressed the OK button to conclude the operation and generate the new image, a RESAMPLE dialogue box flashed about 9 times (presumably one time each for the nine source images??) and I am curious what is being resampled.

Using the crop tool, I laid the border lines along the edges of one of the (nine) images in the resultant image to confirm its dimensions and it appears each panel remains at 2333 pixels in size, so it appears they were not resized in any way.

Just curious.

Thanks,

Marv

[jsachs]

In general, Layout usually needs to resample the images to get them the right size to fit in their panels. Since the last release version of 7.20, I in fact added a 1:1 mode that preserves the sizes on the original images, but I am still working on converting it for version 8.0.

Resampling to the same size as the original is the same as a copy, so there should be no harm done to your images.

[Marpel]

Thanks for the reply. Good to know.

Marv

[Marpel]

Jonathan,

Out of curiosity, I tried two methods of resizing images for the panels.

Each original image is 4666 pixels square. Each of the nine panels is 2333 pixels square.

First I kept the images in their original size, identified each, in turn for its panel, and let the transform resize.

Second, I resized each image (Lanczos 8 x 8), then added to the panels.

I ran both results through Composite > Absolute difference.

I then initiated Levels and Colour and extended the range and found an apparent difference, mostly in the darker tones.

I am wondering if resizing first results in a better image?? And, because this adds another step for each image, if the difference is appreciable or not anyway??

Marv

[jsachs]

I doubt you will be able to see much of a difference.

[Dieter Mayr]

Jonathan, what resampling method is used in Layout ?

Marv, if you want to compare the resizing in Layout with Resize you should use the same resampling-method in both cases.
If different methods are used in Layout and Resize minor differences would be normal.

[jsachs]

Bilinear

[Marpel]

Dieter,

I was not necessarily trying to compare the same re-sampling method used by Layout against Resize, but more was interested in determining if there was a quality difference between the method used by Layout and the Resize re-sampling method I typically use (Lancsoz 8 x 8), and whether the difference in quality was worth the effort of first Resizing to the panel size (given that in the example I was using, I would have to Resize 9 photos before using Layout - a fairly significant added step).

When I did the comparison (using Absolute Difference) I did find a difference, but then realized that it did not indicate which workflow is best quality, only that there is a difference. Hence my question.

Regards,

Marv

[Dieter Mayr]

Marv,

But you can only compare the quality if you know what method is used by Layout.
If your specific image rescaled to your specific size has a (visible) better quality using a resampling method other then bilinear you should probably resize in an extra step.
Expanding the range on a absolute difference image can be dangerous in showing errors one should not care about.
I usually do not expand, if i do not see differences in the absolute difference image it is usually good enough for me.
Expanding can be helpfull to show where the errors are, but as said, i don't use it to determine if I should worry about the errors.

[Marpel]

Dieter,

All good points.

Because I am curious (and I admit, a bit anal!!), I often like to know things like this regardless of whether it has any real practical value.

Cheers,

Marv