Elements - the five down-sampling methods compared

[xpatUSA]

I used a really fancy Siemens Star target image, created by Bart van der Wolf, see here (huge file). It was for lens testing but, 3685 pixels square, seemed ideal for testing re-sampling methods such as bicubic, bilinear, etc. The big file was re-sized from 3685px to 640px for the test. No photographs were involved and no animals were harmed.

Without further ado:









It certainly brings out the aliasing caused by the nearest neighbor algorithm!

In the end, I preferred bilinear in PSE 6 but lately have started to use Lanczos3 in another editor.

[xpatUSA]

In the end, I preferred bilinear in PSE 6 but lately have started to use Lanczos3 in another editor.

Here's a Real World scene at 23% reduction, Lanzcos on right:



Unlike patterned target comparisons, the differences are less obvious but, none-the-less, are discernible to the critical eye. At right, a reduction in moiré on the ribbed insulators especially the inverted chevron artifact on the vertical one. A slight improvement in contrast, see the shadow on the pole.

Not a lot of difference in the Real World but possibly enough to avoid a rejection due to aliasing on a wedding dress or a model's hair?