When reducing the size of an image layer, I am seeing very different quality between regular raster layers and smart object layers.
See this image for an example:
Both versions of the image were resized after duplicating the same source layer.
They were then resized to 20% of their original size.
In the top example, the layer was resized immediately after being duplicated.
In the bottom example the layer was converted to a smart object before resizing.
You will notice that the top version of the image has considerably more artifacting than the bottom.
The quality of the raster layer size reduction also seems to have degraded since CS5.
I am using the bicubic setting in Preferences > General > Image Interpolation.
Switching that option around has made very little difference.
Does anyone have any insight into this?
Are others seeing the same phenomenon?
In CS6, like CS5, the resampling method when transforming a Smart Object is that which is defined in Preferences > General. However, unlike CS5, CS6 has an independent resampling control in the Options bar when a raster layer is being transformed. Looks like you had Nearest Neighbour selected there.
That was exactly the problem.