Smart Object with transform resizing problem

I work in Photoshop for web production all day long and what you are describing sounds very familiar. It is an issue that manifest when a screen is skewed. Photoshop tries to remember every original detail and what ends up in the end is something with too much information and looks jagged. We apply a technique to solve the resolution issue.

"An embeded smart object with a perspective transform get jaggy line after a resize compared to a rasterize layer"

Try doing the following:

• After you do your smart object resize, make is a smart object again. Essentially you are wrapping it inside another smart object.

• Then go inside the new smart object. Inside is where you will adjust the image resolution.

• Using the "Image Size", increase the current resolution by multiplying it by 2. Example, if your image is 72ppi, increase it to 144ppi. (If you are dealing with something thats extremely angled then do this step twice. Example, 72ppi increases to 288ppi.)

• Then save and close the smart object so you are back to the main level of your Photoshop file.

A Smart Object layer has some overhead.  They resize like a Raster layer however, if you may be resizing a raster layer more than once you should convert the raster layer to a smart Object before any resizing because they resize better than raster layer. Every time you resize a raster layer the layer looses some image quality the layer's quality degenerates some.  Though smart Object layers resize like raster layer via interpolation the smart object layer overhead preserves the layer quality.  The Layer pixels are not changed at all. just the rendering for the layers composite rendering is.   This transformation is always done from the layers pixels that where rendered by Photoshop for the object and these pixel are set in concrete they can not be changed.  The layers image quality does not degrade the current layer transform for the document composite may be poor if the transform size change is large.  For example if object is text and when Photoshop rendered pixels for the object text  and shapes become pixels its a raster layer.   If the text characters are rendered with a black line width of one or two pixels resizing the layer  to 33% would mean the text line width would be 1/3 or 2/3 of a pixel the transform will be awful you do not the pixels needed to do a good transform.  Do not crate object that you will transform too small or too big.   There is a range of size transformation  that will work work well with an object. resizing outside that range will produce unacceptable results.

The example 2500x2500 document where  I perspective transformed a smart object layer that contained text then resized then 2500x2500px document to 33% of that.  Resized the smart object not to 33% of the object size it resized it to 15% of its size.  For in the 2500x2500PX document the object was transformed to 47% in the document composite view.  

If you have a high quality object it can tolerate quit a large change in size.  If an image has a large about of detail like text does  you need a lot of pixels to record all the fine detail.  If you are constrain to some number of pixels you can record just so much detail.  The image quality is not as good as it could have been had you more pixels the to store more detail.   A digital image is stored in pixels. 

When you change an image size via interpolation you not changing it size by just changing its DPI resolution your doing it for displays they can not change their DIP their Pixel size is fixed. Your changing the number of pixels you render the image with.    Pixel is where the image details is.   If you reduce the number of Pixels you have to discard details you now have.  If you increase the number of pixels you have to create detail you do not have for the image. When you change an image size by changing the number  of pixels you loose some image quality.  

If you loose too much quality the image rendered will be unacceptable.    If you have a high quality Image it will have a lot of pixels finely recording all the image's detail.   A high quality image can withstand a large range of size changes.  UP or down in size the quality lost will not too much.   On the other hand if you have a lower quality image it will not be able to withstand a large change in size you be discarding too much of the little detail you now have.  If you enlarged you will not be able to create good details for you have too little detail to go on.

IMO you problem is you create low quality objects.

Same test done with the file TestResize.tif - Google Drive with Photoshop CC 2015,5 on another Mac computer leads to the same result.

Cris Cox admited here Photoshop: EPS or PDF Smart objects render badly (with jags) when being scaled or transformed. | Photoshop Family Custom… that a bug like this occured with PDF and EPS 4 years ago.

They say they corrected this but never investigated the general case. And M Cox is unfortunatly not part of Adobe anymore.

Is there someone from support that can help ?

Image size is not a cropping tool. It cans distort an image if the width and height are not constrained linked to each other. So it can not change the image's content it can just distort  the image's composition .  The Image Size Preview is just not showing the whole image because of the size of the preview window the zoom level of the preview image does not fit within the preview window.  Here I show an image size dialog with the preview image zoom set to 100% actual pixels. That number of pixels can not even fit on my display  and surely can  not fit in Image Size preview window. No resize has been set here yet. It is not a bug the zoom level.

As I understand it Smart Object instances are not calculated based on your Image Size Resample settings but based on the General Preferences Image Interpolation setting.

So you might have resampled the plain Layer with Bicubic but the Smart Object with Bicubic Automatic for example.