Render then Export to Image (resolution)

[barkest]

Hi,

I render my image at 10dpi and export it to a jpg at 10dpi

I then export the same 10dpi render to a second jpg at 300dpi.

Everything works as expected in that the second jpg has 30 times more pixels than the first jpg but is exactly the same physical size.

During the export to jpg process those 30 times more pixels came from somewhere.

In PhotoShop the additional pixels that are created automagically are interpolated from the existing pixels using various methods (bicubic for example)

Does anyone know how the extra/few pixels are calculated when I change the export dpi value to be different from the render dpi value?

(My rule of thumb has been to always export the image/pdf at the render dpi)

thank you

[IanH]

Huh?

I'll keep example in inches for simplicity...

Assuming that you have a 1" x 1" image, at 10dpi it will be 10 'dots' x 10 'dots' in size, each 'dot' being 1/10th of an inch in size.

Same 1" x 1" image at 300dpi, will be 300 'dots' by 300 'dots', each dot being 1/300th inch in size.

The extra pixels come from the renderer resolving the model in greater detail.

But both images are physically the same size at 1" x 1", but 10dpi version probably less bytes in size.

Or have I misunderstood something in the question?

[barkest]

Thanks Ian,

must be the way I explained it so I will try again

Sheet layer set at 10dpi but you then export the image at 300 dpi.

You do not get x30 the detail because the detail never existed from the 10dpi render.

You do get x30 pixels (of course) but how are those pixels created or rather 'in what way are they created'?

I assume some kind of very simple interpolation but I do not know what?

hope that makes more sense

[CipesDesign]

That's probably a question for the engineers... I would assume there are various different programming methods... My question to you, because I'm curious: why do you need to know?

[barkest]

Hi Peter

Its a fair question.

I was asked by a student and I explained, with the aid of a 10dpi rendered example, that exporting at 300dpi only takes the existing pixels and averages them to create new pixels so you will have more pixels but a pretty blurry result.

I further explained that the export dpi should therefore match the sheet layer dpi. Of course the comment was that the VW export should default to the sheet layer dpi.

I am assuming my answer is ok (maybe not but it makes sense to me) and then wondered what method of interpolation the export uses

[CipesDesign]

Cool. Thanks for clarifying. Always good to have context!

[Monadnoc]

My understanding is that VW re-renders internally on export to the dpi setting you have chosen (ignores the first render sendings you set), therefore they are new pixels, not interpolated from the 10 dpi original. That's why it takes so long to export an image, and why you can cancel a rendering immediately, then export and it comes out fine. A great time saver for large high res images.

[barkest]

Thanks for the reply

I do not see this behaviour at all.

I have the resolution (export image file) set to 300 and the sheet layer is 10dpi

I have the update visible out of date viewports prior to exporting checked and then I export but as before its 300dpi but interpolated from 10dpi. I just ran a test on this to make sure.

thanks again

[Monadnoc]

You're right, I just tested it myself. So much for that theory. Your advice to always export at the same resolution you render the sheet in is best.

[barkest]

Thanks for looking at it.

To me it confirms my 'guess' that the export at a higher resolution interpolates the pixels based on some kind of averaging (nearest neighbour or similar).

I am not sure how to contact the tech guys to find out the information for this so if anyone knows I would be grateful

thanks

[Monadnoc]

I agree it is interpolating the pixels, probably based on one of the five Photoshop algorithms. And I also agree it is probably nearest neighbor. It would be interesting to find out which algorithm VW is using, just out of curiosity. I've always matched export and sheet layer resolution settings, so this isn't really an issue for me. More just curious.

I wonder if VW does re-render internally when you export from a design layer/saved view, since there is no resolution to set for that. I think I'll go test that later tonight.

From Google:

Photoshop offers five resampling methods. You choose one in the General pane of the Preferences dialog or in the Image Size dialog . The resampling method you choose in the Preferences dialog affects when you resample outside the Image Size dialog, such as when you use the Free Transform tool.

Each method has its strengths and weaknesses for different situations.

1. Nearest Neighbor is the most basic, and it's very fast: To create a new pixel, Photoshop simply looks at the pixel next to it and copies its value. Unfortunately, the results are usually lousy for photographic content. Nearest Neighbor works best when the image has hard, straight edges and when you only want to make the existing pixels look bigger.

2. Bilinear is more complex and produces better quality: The program sets the color or gray value of each pixel according to the pixels surrounding it. Some pictures can be upsampled pretty well with bilinear interpolation, but one of the bicubic options usually looks better.

3. Bicubic creates better effects than Nearest Neighbor or Bilinear. Like Bilinear, it looks at surrounding pixels, but its equation is much more complex and calculation intensive, producing smoother gradations.

4.Bicubic Smoother is specifically designed for upsampling. As its name suggests, it gives a smoother result that handles subsequent sharpening better than the Bicubic sampling method does.

5. Bicubic Sharper is designed for downsampling. It does a better job of preserving detail than the Bicubic method does.

The differences between these resampling methods are often subtle. As a starting point, use Bicubic Smoother for upsampling (but don't expect miracles) and Bicubic Sharper for downsampling, but always test different resampling options when working with critical images.