ADVANCED COMPOSITING & RETOUCHING - Chapter 4, Exercise 1
Why Scaling Down and Back Up Loses Quality on a Normal Layer, But Not a Smart Object
====================================================================================

QUESTION:
Explain what specifically happens to image quality when a normal raster layer
is scaled down and then back up to its original size, and why a Smart Object
avoids that problem.

SOLUTION / EXPLANATION:

A normal raster layer is just a grid of actual pixel values - there's no
memory of what the layer "used to be," only what it currently is. When you
scale that layer down to, say, 20% of its original size, the software has to
throw away roughly 80% of the original pixel information, because a smaller
grid physically cannot hold as much detail as a larger one. The pixels that
get discarded during that downscale are gone permanently - the layer's data
has been irreversibly replaced by a smaller, lower-detail version.

Scaling that same layer back up to 100% doesn't restore the lost detail. The
software can only interpolate (estimate) new pixel values based on the
already-reduced data that remains - it's inventing plausible-looking pixels
to fill the gap, not recovering the original ones. The result is a
noticeably softer, blurrier version of the original, even though the
dimensions match again.

A Smart Object avoids this because it doesn't work by continuously
overwriting a layer's own pixel data with each transform. Instead, it keeps
the original pixel (or vector) data preserved, untouched, inside the Smart
Object container, and every transform you apply - including scaling down and
back up - is stored as an instruction, re-rendered fresh from that original
preserved data each time the Smart Object is displayed or flattened. Scaling
down to 20% and back to 100% on a Smart Object simply means "render the
original data at 100% again," which loses nothing, because the original data
was never actually discarded in the first place.

--------------------------------------------------------------------------
WHY THIS WORKS AS AN ANSWER:
It explains precisely why a normal layer loses information during a downscale
(pixels are physically discarded, not just visually resized) and why an
upscale afterward can only interpolate rather than restore, then contrasts
this with how a Smart Object's preserved-original-data model sidesteps the
problem entirely by never discarding the source data to begin with.