DSLR Photography Forum

Full Version: chemical printing and pixels
You're currently viewing a stripped down version of our content. View the full version with proper formatting.
I get my pictures printed in a little photo shop, where they say they have a "chemical printer". And although I am still not quite sure what that actually means, -and no one even on ST has explained it to me so far! ;-) - the quality of the prints is excellent.

However, the lady in the shop told me, that the edges of my prints would always be cropped by a few mm's. to avoid that, I should size them to the printing size (say 8x10 or 8x12in)
at 320ppi. She actually claimed that this was true for "every professional lab".

Why is that?

And what if I don't actually have enough pixels for 320ppi on an 8x12in picture?
For all my experience, I don't like photoshop to interpolate and resample my pictures.

the other option is, to give them a file of 12.8in x 8.533in at 270ppi. (instead of 300ppi)
(for clarity's sake: 8 X 12 x 320sq = 12.8 x 8.533 x 300sq)

Since I don't know, if chemical printing actually works in dpi, I have been wondering if it is better to resample the pictures in PS or to, so to speak, let the printer do the interpolation.

what do you think?
and why do the edges get trimmed if I size the picture at 300ppi?

uli
Big Grin Until someone who actually understands this comes along, may I suggest a game of cards, bone juggling, or bury the chew?

I could perhaps sing a song, but my voice is a bit ruff, ruff!

Whistling is good. I pretend to know the answer, but just whistle in an innocent, yet knowing way. You conclude that I know the answer..... My pride is intact, and you wait patiently....


OK, by now this thread should have the attention of someone who actually DOES know the answer.

HTH!!!!
I don't think the fact that it's a chemical printer has any bearing on this. The reason part of the picture has to be cropped is that most digital sensors don't naturally produce perfect "standard" print sizes at all sizes .... they'll tend to be a bit over/or under on one side so cropping is necessary to avoid uneven borders on the print. That's why it's best to crop and resize yourself - that way you won't have your composition ruined by the printers. Although 300 dpi is best for quality prints, anything over about 250 should print fine so if I hadn't enough pixels for that I'd always drop the dpi down slightly rather then let software interpolate up, or choose a smaller print size. (That's just me though Smile ).

I tend to set the crop tool to the required print dimensions and dpi to achieve a pleasing crop for the print.

Hope that helps!
Jan

PS Chemical printing just means that the images are printed by the old fashioned chemical method on photopaper in an automated minilab much the way traditional film is processed in those 1 hour shops....Makes no difference to the crop.
Hey Uli... I'm not exactly sure whether she's got her facts mixed up since I've never operated a minilab, but I've got a suspicion that it's the case...

Here's what I think the story should be:

1. Chemical printer: probably just a term for a "lab printer" - as opposed to maybe inkjet or home dye sub? I dunno I'm probably making things up here. Maybe ask her what model machine they've got, and then you can look it up on the net to find out the specs

2. Trimmed edges: I would suspect that most printers automatically size the print so that it is just a little bit larger than the paper itself, to account for misalignment. If they size the print exactly to match and the paper is out by even a quarter of a mm, there would be a horrible white edge which would ruin the photo.

I've noticed that my inkjet printer at home does that (crop edges/oversized print) when I print full bleed to the edges.

3. 320ppi: All commercial printers recommend that you have pics of a certain resolution so that when you enlarge the print, there is actually enough data to provide detail in the print. Otherwise they will have to start interpolating and it will start to look pixelly. 300ppi is what most commercial printers recommend - so 320 sounds reasonable.

4. 8x12 size: just thinking logically about how this could relate to cropped edges. If you take into account #2 above, then say hypothetically if they allowed 2mm for trimming, then 2mm on a 6x4 print would be a significantly larger portion of the image vs 2mm on a 8x12 print. So probably the larger you print, the less you "lose".


I really wouldn't worry too much about the trimmed edges - you probably wouldn't miss it unless you had some detail in the very edge of the photo that you wanted. Maybe try printing one pic with them to see how you like the results before doing a whole job with them.


Anyway, these are just my guesses - I hope they're useful regardless. Big Grin
A bit more info on "chemical printing" taken from an article found via Google:

"Many retail stores print digital images on photo paper using traditional chemical processing in a semi-automated minilab behind the counter. "

I expect that's what she means.

As ST says, they do need to allow for a slight crop to avoid borders, but you should size your prints according to the actual dimensions you want eg: 12" x 8". The machinery then expands the image slightly to ensure a true borderless print at the size you want - my Canon inkjet does the same thing. In effect you lose that tiny bit off the edges, but not enough to make any difference.

Jan
thanks guys,
glad to hear it's common for printers to do this trimming thing, I wouldn't know, because I have never printed any pictures myself.
If what the shop lady told me - and rufus didn't understand - is true, the trimming would be quite significant (0.8inches and 0.5inches on an 8x12in).

Jan, good to hear you feel the same about PS resampling, so I'll try the lower res.
The cannon print manager automatically enlarges prints for borderless printing. Which in effect causes very slight cropping from the edges.
Yup - mine's a HP, and does it when printing borderless...
I have two words for you Uli... "Genuine Fractals".

If the printer prints at 320ppi, then your image *will* get resampled to 320ppi one way or the other. Either you can do it before you give it to the lab, or the lab's computer/printer will do it for you on the way to the printer.

If its something important to be printed (or you are doing it quite large) then I'd prefer to do the upsampling myself where I can see the results and even perhaps apply some more sharpening or something before passing it to the lab. I'm sure some labs would do a good job with the upsampling, but I'm equally sure other labs wouldn't.

There are a heap of different methods for interpolating the extra pixels when upsampling an image. The worst of which is the chunky effect you get by just duplicating pixels, the standard photoshop resampling methods are somewhere in the middle, and then tools such as "Genuine Fractals" are regarded as the best.

Genuine Fractals is a photoshop plugin that allows you to upsample images, maintaining significantly more detail while also antialiasing (smoothing) the results.
Its quite expensive to buy (and most suitable when you need to enlarge an image a LOT), but they have a free trial version you can play with. There might also be similar (but free or at least cheap) alternatives out there too.

Here's their official site:
http://www.ononesoftware.com/detail.php?prodLine_id=2

And here are some article links for some food for thought:
http://www.creativemac.com/HTM/Features/...-page1.htm
http://www.geocities.com/roberthaus/olyt...ample.html

ps: My Canon printer also enlarges prints when doing borderless prints by default. It has a little slidey-bar you can move to set how much it should enlarge, so you can tweak it down to a minimum (or none) if you can manage to put your paper in straight! Wink
numbers are confusing indeed I realize.

the actual pixels I get in the highest quality jpeg from the 350D is 3456x2304.
that gives a picture of 11.5 x 7.7 inches at 300ppi OR
a picture of 12 x 8 inches at 288ppi.

My original question was, wheather it was better to interpolate in photoshop or to let the printer do it (which I guess also depends on the printer and its software).

From experience in my PhD where I was doing a lot of microscopy and digital capturing, I remember the advice to rather print a lower res file than to resample in PS,
however, those pictures were rather small, and the printer was high end.

Jan seems to agree with that, which I found comforting.

To complicate isssues, the store lady wanted a file of 12 x 8 inches at 320ppi, which if you think about it now, does not make any sense!
If the printer really enlarges the file slightly to fill the paper, than she should have asked me for a slightly smaller file, not an effectively larger one.
that's still a mysterium for me.

I wouldn't mind asking her more details if she wasn't a) chineses, and b) already annoyed about me asking so much Tongue
I always give them a small sample of a picture to print at 4x6 before I print a large thing.

genuine fractals would maybe help, but I don't think it's worth it for just my pictures. If you look at the samples on the geocities link though, photoshop seems to do okISH even at resampling a picture to 2.4 times its resolution (!! which is WAAAY more than what I would need), although those pics are quite small.

I guess, I'll just try both........ will let you know

:/

uli Wink
Hey again Uli,

Ok.. taking Genuine Fractals out of the equation as impractical... I'd still imagine you'd be better off resampling in photoshop than letting the printer do it.

In much the same way as photoshop does a better job of colour management than pretty much all printer drivers, and does a better job of RAW conversion than your camera will... so too I'd trust its resampling algorhythyms more than a printer or printer driver.
Photoshop's in-built functions might not be the absolute best (just like its RAW conversion process isn't) but they are still professional-quality and certainly very good, and it leaves the quality control in your hands, not at the mercy of the lab printer (or worse, the lab operator who might be different from day to day). If nothing else it should guarantee a level of consistency.

But at the end of the day, with the small amount of upsampling involved and the high-resolution of the original in your example (maybe 10% upsampling??? a 3456x2304 image printed at 12x8"), I'm sure there will be no detectable difference between *any* of the methods discussed. I've printed many such shots this size and you simply can't make out individual pixels. I think 300ppi is supposed to be the point where the human eye can no longer distinguish detail (don't quote me on that, it just rings a bell), and considering the differences will at most be very subtle, I really don't think there's much to worry about at these kind of print sizes.

If you are planning on larger prints (24x16" or bigger) then these resampling issues certainly become more important, but personally at the moment I'd be more concerned about colour management than resolution with regard to your prints, as that's an area which can really cause some disappointment when things aren't right. Particularly if the screen you use to edit images on isn't calibrated.

Cheers
Adrian
Hey Adrian,

have you ever printed anything larger than 8x12in from the 350D? If so, how did it turn out?
I just kind of assumed that the resolution wasn't enough for that, because for anything beyond 8x12 resampling is definitely required.
As you say, it is not actually a crucial issue for what I am doing right now, and as you can see, 2 people (Jan and you) have two different ways of doing it....
I just like to be accurate, and consistent, and I like to understand why I do things the way I do them.

As for the colors, that's why I give them a 4x6 crop from my pictures before I order an 8x12 print, costs me 29cents and I know what I get. Smile

Cheers, Uli
I'm not Adrian, but I've printed 20x30" images from my 6mp Canon 10D, and they are as good as (and in some cases much better than) similar-sized prints from 35mm negative film.

I resize in photoshop in 10% increments until I reach the size I want at 300ppi, alternating between "bicubic smoother" and regular "bicubic" interpolation - but never "bicubic sharper." Then I sharpen with Focal Blade, a worthwhile investment IMHO (see the review on this site.)

Or take a look at Fred Miranda's Stair Interpolation and Intellisharpen plugins. They are good values and used by many pros.
http://www.fredmiranda.com/software/

BTW, I believe the Noritsu printers are optimized for 320ppi, as are the Chromiras. If that's what your lab recommends, then that's what you should give 'em.
Thanks slejhammer!

I'm going to the shop today anyway, so will ask them what model their printer acutally is.

I only have version 5 of PS, is that why I only have the options of
bicubic
nearest neighbor or
bilinear ?

I always do bicubic......
Bicubic should be fine. The 10% increment method (aka "stair interpolation") was made for that. Some pros (PS guru Ben Willimore) claim that with bicubic smoother you don't need to do 10% increments anymore, and can resize all at once. I'm a bit dubious ... anyway, "smoother" is a little better than plain ol' bicubic, but you'll be fine without it.

Oh, one other thing that seems to help image quality for really big enlargements: shoot in RAW and convert to 16-bit TIFF, and then do your resizing in 10% increments. I typically don't find much advantage to 16-bit processing, but for enlargements it really seems to make a positive difference, at least to my eyes. (The difference is subtle on screen, but the final prints seem more vibrant and crisp.) The drawback, of course, is huge file sizes. Not sure if PS5 supports 16-bit though.
yep, it does. thanks for that bit. will try.
Hey guys,

Uli, I've printed a small handfull of 350D images at A2 (approx 24x16"), and they have been lovely and sharp (but had been enlarged using Genuine Fractals).

Peto.. I'm quite curious about the "stair interpolation" method. I followed the link and had a bit of a read through Fred Miranda's info (http://www.fredmiranda.com/SI/)... and was left scratching my head a bit. I think in the examples he showed, the SI method looked worse than the standard photoshop bicubic examples (to my eyes anyway), and I don't know what happened with those Genuine Fractals examples - they were the worst looking GF samples I've ever seen.

Also, as a software developer who's developed a number of interpolation algorhythms for various things from graphics and audio to blood lactate curves, at a glance my experience (without knowing the details behind photoshop's bicubic resampling code) is telling me the SI method *should* be worse because it compounds all the errors that are introduced with each iteration of the process. Its a little bit like how when you repeatedly load, change and save the same JPG file it will progressively lose quality due to the lossy JPG compression method. Whereas if you do all your edits at once and just save JPG once at the end, you aren't compounding the problems caused by compression artifacts.

Some of these errors (the rounding errors which cause posterisation) will be offset by the fact you can perform SI in 16-bit mode, but this won't prevent other issues which may cause noise and aliasing or a loss of detail.

But reality always beats theory... so if the results really are better, then I stand corrected. Smile
I also have Genuine Fractals and it is very good for uprezing and downrezing photos. It is painfullt slow though, and if I click on one of the icons for a Genuine Fractals file it crashes Windows Explorer.

The gradual resizing in 10% increments that Mitch describes works very well - you can resize almost indefintely using that technique. Best to not sharpen before resizing though - sharpen only the final output for best results.