More help please, still got trouble with file size from my new 5D

Hi everyone, can I ask for more help please, Cem your suggestion re changing to 16-bit, worked perfectly, but I have a complicating factor in that an Agency that I work for require files to be 8 bit how do I resolve this issue? I have previously interpolated in Photoshop by amending the "pixel dimensions" by going into Image > Image Size> and then amending the size until the desired file size is obtained. Is it Ok to do it this way?

in cs2 - image -mode

Hi Jan,

(Being short, 'cos I know you want to tear into it ;-)

probably best to work in 16 bit, then change to 8 bit for your agency.

You can not adjust bit depth, image size, number of bits, dpi, file size, independently - they are all related.

Exactly what is the agency requiring? If you ask half a question, you get maybe, half an answer ;-)

Best wishes,

Ray

Hi Ray

I am trying to create a 75MB file from new 5D but needs to be 8 bit not 16 bit, when I revert to 8 bit the file size halves, am I correct in assuming I can change pixel dimensions until I achieve the desired file size?

Many photo labs still can't deal with 16 bit images either. Just open the image in Photoshop and go to IMAGE>MODE and click 8 bit. You won't see any difference on screen, but save the 8 bit file with a different name so you still have the 16 bit file as backup.

Hi Steve

The problem is that when I go back to 8 bit the file size halves back to 36mb, I need to supply a 72 - 75 mb file at 8 bit, so I have amended pixel dimensions to approx 6225 x 4150 8 bit to achieve 72mb file approx. Is it incorrect to interpolate this way?

I am trying to create a 75MB file from new 5D but needs to be 8 bit not 16 bit, when I revert to 8 bit the file size halves, am I correct in assuming I can change pixel dimensions until I achieve the desired file size?

Click to expand...

Janet,

A 36Mb 8-bit file will print at the same size as a 72Mb 16-bit file, all other things being equal, so you work in 16-bit and then as a final step convert to 8-bit for printing. No need to worry.

number of bits, dpi, file size,

Click to expand...

Ray,

Unless I'm very much mistaken the dpi setting has no effect on the file size. It's a tag associated with the file which may or may not be used by the printer driver - probably not.

And to be pedantic, it should be ppi, not dpi, which refers only to the actual printer.

Regards,

Stuart

Hi Jan,

There are various 'upsizing algorithms'. Basically they have to generate pixels, based on ones near by. I am not sure if the ps version is good, better or worse than others. Can you not send the file to the agency, and ask if they would say if it was no use? I think it will be OK.

Best wishes,

Ray

Hi Ray

File has to be 8 bit, they will reject it if I send it as 16 bit. Thanks for your help though, I'll look into this further.

Jan

Hi Stuart,

I was trying to illustrate that a number of 'dimensions' that are bandied about depend on each other, and it is quite possible for a specification to be none achievable. I think that the agency is asking for an 8 bit file of 72MB. In effect, they would be safer asking for a 36MB file or better still 16 bit file. A 72MB upsized from 36MB is more useless than the original 36MB file imho - its a version of saying my p&S with 8MB is better than your pro camera with 5MB - quantity, not quality.

Best wishes,

Ray

Hi Jan:

1) Work on your image at native size and in 16-bit to satisfaction (I also recommend at least working in Adobe RGB or even Prophoto RGB with 16-bit files).
2) Now "Uprez" the image, still in 16-bit -- to DOUBLE your desired file size. I recommend Photoshop because it works well, and still do it the same way I describe in this article: http://www.outbackphoto.com/workflow/wf_60/essay.html. (I action this, but Uwe wrote a plug-in version you can download and try for free, link at the end of the article.)
3) Change the mode back to 8-bits per channel, which will cut your file size in half, and down to your original targeted size -- in CS this is: Image>Mode>8-bits/channel
4) Final step, convert to (not assign) whatever color-space your editor wants/needs the file in.

Cheers,

Hi Jack

I'll try this, thank you very much.

Hi Jan,

I guess you can't read between my lines. ;-(

1) do your editing in 16bit.
2) convert to 8 bit - cs2 - image -mode
(your image file will be half size, in 8 bit e.g. 36MB, if you were to save it)
3) resize to twice its linear size - cs2 - image - image size - width and height
(if it's say 800*600 make it 1600*1200 pixels)
4) save it - that should be it, assuming your original 16bit file is 72mB or so
then this saved 8bit file will be 72MB

I was saying that I was uncertain if the cs2 upsizing algorithm was an unknown quantity,
in comparison with others. It is probably good enough. I was suggesting you send the 8 bit file. If you are pc based, then get qimage.

Best wishes,

Ray

Hi Ray

Thank you for this, best wishes

Jack_Flesher said:

Hi Jan:

1) Work on your image at native size and in 16-bit to satisfaction (I also recommend at least working in Adobe RGB or even Prophoto RGB with 16-bit files).
2) Now "Uprez" the image, still in 16-bit -- to DOUBLE your desired file size. I recommend Photoshop because it works well, and still do it the same way I describe in this article: http://www.outbackphoto.com/workflow/wf_60/essay.html. (I action this, but Uwe wrote a plug-in version you can download and try for free, link at the end of the article.)
3) Change the mode back to 8-bits per channel, which will cut your file size in half, and down to your original targeted size -- in CS this is: Image>Mode>8-bits/channel
4) Final step, convert to (not assign) whatever color-space your editor wants/needs the file in.

Cheers,

Click to expand...

Jan,

Jack has given great advice. I recommend that you read this article fully to understand more. IMO, the method of Jack will give you the best results. File size goes like this during this process:
Step 1: 72.8 MB (16-bit ProPhoto or Adobe RGB file at a resolution of 4368 x 2912 pixels x 3 x 2 = 72.8 MB)
Step 2: 145.6 MB due to uprezzing from 4368 x 2912 to 6177 x 4118 pixels
Step 3: 72.8 MB again, due to conversion from 16-bits to 8-bits at a resolution of 6177 x 4118 pixels.

This will assure the highest achievable quality given the limitation of the sensor size of 5D. If you really want to have 72 MB pixels without uprezzing, you'll have to either buy a 24 MP SLR (which does not exist) or a MF digital back or (drum) scan film material.


Cheers,

Ray West said:

Hi Jan,

I guess you can't read between my lines. ;-(

1) do your editing in 16bit.
2) convert to 8 bit - cs2 - image -mode
(your image file will be half size, in 8 bit e.g. 36MB, if you were to save it)
3) resize to twice its linear size - cs2 - image - image size - width and height
(if it's say 800*600 make it 1600*1200 pixels)
4) save it - that should be it, assuming your original 16bit file is 72mB or so
then this saved 8bit file will be 72MB

Click to expand...

Ray:

IMHO, it's best to resize in 16-bit if your machine has enough horsepower -- the interpolation done, especially when going UP in size, can use the extra colors available in 16-bit to help reduce the possibility of banding in large, even-toned gradient areas like skies... To make this more clear, think of 16-bit as offering and additional 1024 individual color-points between each neighboring color-point that is available in 8-bit...

Cheers,

Hi Cem, Jack, Ray & everyone who has helped

I think I'm all sorted out with what I need to do now, thank you all so much for your help. My understanding of interpolation is now greatly improved. I hope to be able to get on with some serious photography in the next few days and will show you all the results. Luckily I'm much better with a camera than a computer!!

Once again thanks for all your help, it's very much appreciated.

Hi Jack,

I hadn't realized you'd posted b4 my reply. You are quite right in what you say. However, as you mentioned re the pc speed/mem size issue, and my take on perhaps the arbitrary nature of the agency requirements, at the end of the day, it may not matter, but Janet may as well do the best possible job.


- so Janet, do as Jack says!

I spent a lot of time some months ago, with qimage, interesting re. the different resize algorithms. I've forgotten it all now, except slow is best ;-)

Next year. the same agency will be wanting 200MB files, I expect.

Best wishes,

Ray

Ray West said:

...However, as you mentioned re the pc speed/mem size issue, and my take on perhaps the arbitrary nature of the agency requirements, at the end of the day, it may not matter, but Janet may as well do the best possible job....
...
Next year. the same agency will be wanting 200MB files, I expect.
...

Click to expand...

How so true! This is a ridiculous exercise since uprezzing a 12 MP image to 24 MP is never going to be equivalent to having a "real" 24 MP image to start with. A totally outdated and illogical requirement, if you ask my opinion.

BTW, Ray's method vs Jack's is actually a matter of splitting ends. Both will work just as well, I guess. I prefer Jack's due to the possible posterization issues he has already mentioned.

Cheers,

janet Smith said:

Hi Ray

I am trying to create a 75MB file from new 5D but needs to be 8 bit not 16 bit, when I revert to 8 bit the file size halves, am I correct in assuming I can change pixel dimensions until I achieve the desired file size?

Click to expand...

Hi Janet!

Basically working in 16bit is always better. But when it comes to offset-printing each photo has to be CMYK - and that´s 8bit/chan. Seperating cmyk from 16bit RGB profits from a great spread colour-space as possible.
Printing is always in 8bit/chan, that means 256 steps in gray/chan.

In fact i give away RGB-files in 16bit-TIFF and the agency´s pre-press people make seperations they like using their specialized profiles which are matched to their machines.
That´s the way all professional advertising agencies or magazines act.

Important is to use the correct colour-workflow with an optimal rendering intend to change from 16bit to 8bit - which is barely visible, but essential when working the files with contrast, colour and so on.
I´m making composings very often and it´s very important to have the same colour-spaces to get proper results. So that´s a great theme to me and i tested lots of ways. Fact is: experienced professionals stay with 16bit as long as possible and make a change just before finishing a work and give it to the client or give the client the 16bit-TIFFs - which i learned as the best and most safe way.

btw.: a file can´t be halve the size by just switching from 16bit to 8 bit! Stay with TIFF all the time and use LZW as a (lossless) compression.

best, Klaus

Klaus Esser said:

Hi Janet!

Basically working in 16bit is always better. But when it comes to offset-printing each photo has to be CMYK - and that´s 8bit/chan. Seperating cmyk from 16bit RGB profits from a great spread colour-space as possible.
Printing is always in 8bit/chan, that means 256 steps in gray/chan.

In fact i give away RGB-files in 16bit-TIFF and the agency´s pre-press people make seperations they like using their specialized profiles which are matched to their machines.
That´s the way all professional advertising agencies or magazines act.

Click to expand...

Janet,

This is good advice. Also look to Nicolas Claris. He is so careful and experienced with CMYK. He works closely with the printer. He can help you with anything!

Now what do stock agencies or art directors mean when they specify a 75MB file? Is that meant 75MB in 16 BIT or not?

Asher

Hi Klaus,

Although the reasons you've given to use 16-bit are valid, I think you might have missed the point here. Jan's problem is not the file size, which can be easily doubled by switching from 8-bit to 16-bits. Neither is it about the prepress people or colour managed workflow. Her problem is that her stock agency wants her to submit files that are at least 72 MB in uncompressed 8-bit format.

Klaus Esser said:

...
btw.: a file can´t be halve the size by just switching from 16bit to 8 bit! Stay with TIFF all the time and use LZW as a (lossless) compression.

Click to expand...

I beg to differ with this statement. For an uncompressed TIFF file, switching from 16-bit to 8-bit will halve the file size, ignoring the odd control bytes used for tiff file format itself. Just to test it, create a 8-bit 500x500 TIFF file in PS and save without layers and without compression as a TIFF file. You'll get a file of around 751 KB. I.e. 500x500x3=750000 plus the control bytes. Next, change the mode of the same file to 16-bit and save a copy as TIFF, again without layers and without compression. You'll get a file with the size of 1,483 KB.

Regards,

And LZW compression on 16-bit TIFF photographs (from RAW files) almost always increases* the file size.

*Because compression works on repeated patterns, which are lacking in these types of images.

With all due respect to Ms. Smith, I must be missing something.

Native uncompressed output of the 5D is 12 bits/pixel (1.5 bytes) x 12 MPixel = 18 MBytes. Any file larger than this must have data replicated, added, interpolated, or fabricated!

It seems curious to me that the goal is to generate 4 times this amount of data, while simultaneously reducing the bit depth to 8 bits/pixel.

Even if we assume (generously, but incorrectly) that each pixel on the sensor is represented by 8 bits per color, that yields no more than 36 MBytes.

What is the intended use of these images? I don't understand how increasing the image file size is of any benefit when you can't possibly increase the camera's native resolution.

Chuck Fry said:

With all due respect to Ms. Smith, I must be missing something.

Native uncompressed output of the 5D is 12 bits/pixel (1.5 bytes) x 12 MPixel = 18 MBytes. Any file larger than this must have data replicated, added, interpolated, or fabricated!

It seems curious to me that the goal is to generate 4 times this amount of data, while simultaneously reducing the bit depth to 8 bits/pixel.

Even if we assume (generously, but incorrectly) that each pixel on the sensor is represented by 8 bits per color, that yields no more than 36 MBytes.

What is the intended use of these images? I don't understand how increasing the image file size is of any benefit when you can't possibly increase the camera's native resolution.

Click to expand...

You are indeed missing something Chuck, please read my post immediately before this one.
We are talking about the TIFF file generated by converting the RAW output of the 5D, which will give a file size of 36 MB as you yourself have also concluded. The TIFF file has a resolution of 4368 x 2912 pixels x 3 x 1 = 36.4 MB

The goal of Jan is: submit uncompressed, 8-bit TIFF files to a stock library at the minimum allowed size of 72 MB!

The actual/IQ benefit gained by doubling the camera native resolution in order to increase the file size from 36 MB to 72 MB: zilch, nada, zero

The practical benefit gained: the stock library accepts the images of Jan instead of simply rejecting them!

I hope that this clarifies things for once and all .

Cheers,

"I hope that this clarifies things for once and all ."

I missed that thing with the stock-agency. What i don´t miss: it doesn´t make sense

best, Klaus

Klaus Esser said:

"...What i don´t miss: it doesn´t make sense

Click to expand...

I agree 100%! As I wrote before:

Cem Usakligil said:

How so true! This is a ridiculous exercise since uprezzing a 12 MP image to 24 MP is never going to be equivalent to having a "real" 24 MP image to start with. A totally outdated and illogical requirement, if you ask my opinion....

Click to expand...

It is utter madness, really .

Cheers,

Cem Usakligil said:

It is utter madness, really .

Click to expand...

It is madness, and if that's what the agency wants, it's what they'll get. There's little use in fighting a paying customer ... Don Quichote and windmills come to mind.

Bart

Bonjour Janet

First ask your agency if they want RGB or CMYK image.
A 30Mb RGB image will "weight" 40 Mb when converted to CMYK. (4 channels instead of 3)

8 bit is a question of "quantification" of the pixels, it is not related with the size of image.

If you want to be precise about size (area):
don't talk 8 or 16 bits (but work in 16 and deliver 8bit)
specify xx inches/cm by xx inches/cm @ xx dpi
The dpi is somewhat the "depth" of your image which is made, whatever the DPI with length in pixels by width in pixels. NOTHING else.
DPI will let a printer rip to spread your pixels into a certain area, look just hereunder, this is the same unresized IDs MkII image:

Look, whatever the # of DPI you specify, as resize is unchecked, the # of pixels remain the same BUT the area changes.



and here is the same with the same image in CMYK:




The debate about what your agency ask is IMHO not the point here (though, I agree there's a lot to say). I assume that by agency you mean stock agency. If it is an advertising agency this is completly different.

If you want to be precise about size 2:
an 1Ds Mk II image is 4992x3328 pixels = 47,5 Mb and that is closed to A3 that is a standard for your agency.
Most of agencies do ask 50 Mb ORIGINAL/UNRESIZED file and this is completely normal as they whish to be able to sell double spread A4 page images to their clients.

It would be completely foollish to systematically increase your 5D files in PS to 50 Mb. Some will print gorgeous, some not, they need to be tack sharp with a very very good "piqué" before any software sharpening.

Sorry but your 5D is simply not able to produce reliable A3 pages for each shot. But your 5D is certainly able to produce magnifiicent images to sell to magazines. Individually.

Photographers like Bejamin Kanarek do work with 10 Mpix camera such as the Pentax K10D (or many photogs with Canon 1D series and Nikons as well) but they deal directly with magazines. Not with agencies.
Stock agencies don't want to take the risk to store/sell small images, they have a standard: A3 @ 300 dpi.

If you deal with an advertising agency, you'll have to show and prove them that your files are good enough for their need....

Hope it helps!
PS (RGB and CMYK color spaces are another story that we could eventually debate in another topic; what color space do you send to your clients? (printers, agencies, stock agencies graphists etc...)

Hi Nicolas!

Good explanation! Absolutely hits the point.

I´m glad you didn´t do it in those other language you used in the other thread . .

best, Klaus