Same pixels, more megabytes - huh?

[jeremy_horowitz]

I wanted to post this question because I was recently in a situation

that made no sense to me.

 

I was at The Icon lab in L.A. and asked them about their negative

scanning services. The lady with whom I spoke (who I learned is the

head of their digital department) said that for their low-

level/rudimentary "crystal" scans, they had three levels, something

like 5-10MB, 10-20MB and 30+MB. They had corresponding costs, but

when I asked her what the resolution and/or size of the scans would

be, she said 300dpi at 100%.

 

I asked her: If the attributes of the scan are the same at each

level, what accounts for the bigger file? She didn't have an answer

except that more megabytes is better.

 

Am I missing something here? Theoretically, if you get a 5MB file

(for argument's sake) from a 300dpi/100% scan of a negative, a

bigger file would be the result of either a higher resolution scan

or higher magnification (say 600dpi/100% or 300dpi/500%)?

 

When I spoke with the people at Samy's Camera, another shop here in

L.A., they made much more sense, saying that one of their services

is to scan a negative at something like 1000dpi/100%, and you can

get a fairly large print out of that because even though the

enlarging reduces the resolution, it still stays at 300dpi or above

until you get to 12x16 (according to Samy's example).

 

So -- on a 1:1 basis (no variations in file processing, compression,

etc.), is it possible to have two negative scans at exactly the same

settings, yet have one scan yield a file that is many times larger

that the output of the other scan?

 

Color me a bit confused.

 

-Jeremy

[frank_skomial]

Your discussions with the Lab lady missed any real explanation of the scanning process, and format or the quality of the saved files.

 

If you want to discuss such matters, you need to find commond grounds for discussion. I think the lady realized that you do not have sufficient depth of understanding the scanning process. I believe she did not have time to explain better, and gave you short and correct answer.

[kelly_flanigan1]

Demand the lab give you an answer in X by Y pixels for each service level. This is what really matters.

[marshall]

I'd go with what Kelly is saying to start with. If the same negative produces file sizes that are dramatically different, <em>something</em> in the settings was different. That said, it's possible to scan at the same resolution and get different file sizes by using different bit depths (8 vs 16), and it's possible to get different file sizes just because of the file compression if they are saved as jpegs (and you should be getting TIFFs). Ideally, though, you need the information about what the optical scan resolution is. It isn't 300dpi - usually people use "300dpi" to describe a print setting. Scanning a negative at 300dpi doesn't really give you that much.

[bill_t__new_mexico_]

For exactly the same pixel dimensions and bit depth, compressed file formats like .jpg can be many 10's of times smaller than an uncompressed .tif or a losslessly compressed .psd Photoshop file. Files scanned in formats that allow 16 bit depth (this is rare) will be at least twice the size of a files scanned at 8 bit depth.

 

You can save yourself some confusion by specifying the pixel dimensions of the output file you're after. This sidesteps many confusing issues, such as what constitutes "100%." If you want to print an image at 13" x 19", ask for a scan or approximately 3900 pixels by 5700 pixels, a huge file! That assumes 300 DPI (dots or pixel per inch) is a reasonable printing resolution, which it is since this approximates the practical on-paper resolution capability of most inkjets. 300 pixels/inch * 13 inches = 3900 DPI in that dimension. Well, 450+ DPI might have minor advantages, I know somebody will point this out, but 300 should do the job in almost any case.

 

.jpg files with the minimum compression settings (such as Photoshop Quality "12") will for all practical purposes print as well as much larger uncompressed files, while still being dramatically smaller in file size.

 

BTW, a 35mm slide or negative scanned at 300 DPI "on the negative" will look pathetic if printed. The *minimum* reasonable pixel dimensions for scanning a 35mm piece of film for printing is about 2000 dot per inch of the original piece of film, or about 2000 x 3000 dots or pixels in total for 35mm. For 4x5, you would want 8000 x 10000 pixels to start to do justice to a technically good transparency or negative.

[oceanphysics]

300 dpi is an output resolution to a printer. Nobody scans at anything like 300 dpi, and if you did it would give you only about a 150KB file.

 

Probably she meant that the scanner tags the file for print output at 300 dpi. This can be easily changed, so it means essentially nothing.

 

I have no idea what you're doing with the percentages behind your dpi figures, by the way. It doesn't mean a thing to me. 300 dpi is 300 dpi.

[kelly_flanigan1]

One of our early transparency flatbeds was a 600 dpi class device that cost as much as a new Hasselblad kit. A later one was a 1200 dpi class device, and it too cost more than a Hasselblad kit. At that time there were some amateur scan wands and some really expense 300 dpi class flatbeds, that did NOT do a transparency scan.

[mel_booth]

It is clear to me from looking at the answers posted on this subject and on other links also that most people do not understand resolutions and file sizes.

There are no grey areas,just four things that define the file size of the scan.1; area of the original,2;the bit depth,3;the PPI of the scan,4;the number of colour channels (RGB *3)(CMYK *4).There are so many points to make about this subject that people have written whole books on it.

Ocean's comments on this are a bit off,a 300 dpi (PPI)scan of a 35 mm trani. in RGB gives you 352 KB.Using percentages is also valid,imagine placeing a 35mm film strip in an enlarger then blowing it up to a 10x6 image (approx 700 %)then scanning that at the same PPI you will give you over 15 MB.

Back to the original question, most pro labs use lazor drum scanners that can be focused via high quality lenses to anywhere on the film effectively enlarging it, if say the PPI was 600 and you set your output to 300 DPI it will be twice the size without loss of quality.

Don't forget that a pro lab will be scanning from 35mm film up to 10in x 8in film the later will take significantly longer to scan creating a huge file size and therefore a higher cost.

Finaly if the lab saves the file in JPEG format (smaller file size)it should have no relatioship to the cost whatsoever as there is no cost involved to the lab.NB.if they save your file as a EPS it will increase the file size whithout changing the image resolution and should cost you no extra.

Sorry to go on so far but it is a complex subject.

[blumesan]

It's not such a complicated subject, but it gets messy when one confuses or mixes two issues: (a) scanning a piece of film and (b) making a print from the scanned image. Let's stick with the first since this is what the OP asked about.

 

When a lab expresses their scanning resolution in "dpi at X%", find another lab. Scan resolution is expressed as pixels per inch (ppi). The inch dimension refers to the original film (negative or positive); in the case of 35mm thats approx. 1.0 x 1.5 inches. For an even moderately good scan (capable of producing a 4x6" print) the minimum requirement is approx. 1200 ppi. More is better and up to 4000 ppi is fairly common practice.

 

Resulting file size of a scan is easily calculated:

width pixels x length pixels x bytes per pixel = tiff file size.

Bytes per pixel = 3 (for RGB at 24 bit depth) or 6 (48 bit depth).

 

e.g. a 35mm slide scanned at 1200ppi, 24 bit depth gives a file of 1200 x 1800 x 3 = 6,480,000 bytes or 6.18MB.

 

 

Once the scan is done the lab can convert the file to any format you wish. Generally they will provide you with a jpeg file which has been compressed to save space, but you should also receive the original scan output (usually a tiff file) so that you can do your own post processing of the image on the uncompressed image data.

[kelly_flanigan1]

Mel; this subject is real easy, and NOT really that complex at all. The lab is basically being a jackass not to mention what the image size is in pixels, bit depth and resulting file size. Folks did this BEFORE Photoshop came out. A X by Y pixel image at 8bits of RBG was the most common why to express a color scanned image even before photoshop came out. Some of us used photostyler and other ancient programs too. By not being really clear with what scan output one is getting; the lab must be run by some greenhorn newbies with that have lying salemens DNA. <BR><BR>Jermey should find somebody at his targeted lab that has been around the block awhile, that can give a clear answer for each of the labs scan services. At our shop we we be out of business if we have whaffled answers on scan resolutions to customers. The entire subject is real easy, old as the wind, and extremely simple. The entire math required is about the 4th grade level; ie mulitplication and division. If the lab compresses the images the uncompressed file sizes need to be known for each service too. <BR><BR>Just mentioning "dpi" is abit not enough real info. Long ago drum scans were only 600 lines per inch; and high end 2 thousand dollar flatbed transparency units were only 600 dpi. For a 4x5 of 5x7 transparency at that time, a 600dpi scan was often used in the heat of battle for alot of printing work. The 1200 dpi unit we bought cost then about 2 grand, and actually killed of alot of outsourcing of those 600 lpi drum scans. <BR><BR>

[kelly_flanigan1]

If one takes a typical customers input for a scan job most of the real usefull info is in the first "1200ppi" ; some is in the next "1200ppi"; and very little is in the next "1200ppi". Thus what happens is that with high end scanner, the file size grows alot quicker than the real info captured from a typical customers image. This means one can scan an image with a "4000 dpi class" film scanner; and either do another scan at 2000; and often much improvement. All it takes is a micro tad of blur, missfocus, super duper fast film and one dosent often require a high end scan. The general lay public seems to somehow think a drum scan will magically pull out details, that often dont really exist. Part of a lab or prints shops job is to steer csutomers on what is good enough. Getting a highend scan from an average negative often just makes a large file, cost more, and really doesnt pull any really extra info. <BR><BR>Many times lower end scans that are done while processing are good enough to cull out the average shots, and steer the high end scans to images that can support it.<BR><BR>

[mel_booth]

Kelly thanks for your response,my view that this is a complex subject is based on my own one to one decusions with many people and my amazement of how they still dont quite know what I am talking about.It realy is simple when you do understand it but I have learned not to assume that everyone has the same knowlage.I will bet there are quite a few people reading your comments won't have a clue what you mean.Mike also thinks it isn't a complicated subject yet he has got his display colour space mixed up with sampling BIT depth.When people refer to 24 bit they usually mean the viewing colour space ie 24 bit colour = 8bit red:8bit green:8bit blue,8x3=24.However when sampling is involved eg. scanner,camera,it relates to converting analogue into digital.The 24 bit caculation in Mikes example is actualy only 8 bit.I know of no capture device in the world that can capture 24 bit let alone 48 bit.If you were to buy the latest state of the art camera it would almost certainly be 16 bit approx. 65,500 x 65,500 x 65,500 colours.

[blumesan]

Mel,

I don't know how to phrase a polite response except to suggest that you read my previous post again. It is absolutely correct in every respect. In no way have I confused the monitor color space with the data gathered by a scanner. Most of the better desktop film scanners (and certainly all commercial scanner) have the ability to convert the analog data to 14 bits per color. This data is then padded or expanded to 16 bits (2 bytes). Three colors (RGB) x 16 bits = 48 bits = 6 bytes.

[mel_booth]

Hi Mike,

I am sure we both know what we mean there is just a mix up in our terminology.When I

talk about bit depth I am refering to input,if for instance I capture at 8 bit depth and saved

in RGB then of course 8*3=24,I would say I have 24 bit image.However using your

example of scanning a 35 mm slide @1200PPI I would describe that as scanning @ 8 bit to

avoid confusion,otherwise try explaining why, when setting up a new document in

Photoshop the same 1200PPI scan in a 16 bit space will give you twice the file size as one

you describe as 24 bit.You will say of course that this is 48 bit but there is no 48 bit in

Photoshop! can get confusing for first time users.I misunderstood you and I have been

doing this for over 12 years.Maybe this is a subject easier to understand than to explain?

Cheers.

[blumesan]

Hi Mel,

 

First I think part of the confusion is caused by the inconsistency in the terminology used to describe color information depth. Sometimes it is expressed in BITS per pixel for each color, and at other times as total bits per pixel for all 3 (or 4) colors. (And to confuse matters even further, some insist on expressing this data as BYTES per pixel.)

 

Thus an "ordinary" image file with 8 bits of data for each of 3 colors may be described as an "8 bit image" or a file having a bit depth of 24. A so called "high bit" image contains 16 bits of data for each color, and may be described as a "16 bit image" or having a bit depth of 48. By the way, beginning with Photoshop CS (ver 8) almost every function is capable of working with 16 bit images (48 bit color depth) and this capability has been further extended in PS/CS2.

 

Rather than being confusing, this information is directly pertinent to the original question, explaining, in part, why files with the same number of pixels contain different number of megabytes.