Tips for extreme file size reduction?

B
Posted By
bhilton665
Nov 1, 2003
Views
702
Replies
9
Status
Closed
When I take large files from medium format film scans and shrink them to viewable jpeg or thumbnail sizes I lose a lot of the inherent image quality. This is taking files that might start out at say 11,000 x 8,800 pixels and reducing to 600 pixels wide for a web image or 90 pixels wide for a thumb.

I’ve tried the one step approach and also have written actions to do it in 50% increments and in 93% increments, which look better for most images. I’ve also tried re-scanning at 1,000 ppi instead of 4,000 ppi and this works well too, but I want to avoid scanning twice at all costs.

I have two questions …

1) Anyone have any other magic numbers besides 50% and 93% for something like this? Whipping up the action is pretty trivial so I can try different values but life is too short to try all of them :).

2) Anyone aware of a web site that has studied this in some detail in a systematic way? Got a link?

Thanks for any help.

Bill

Master Retouching Hair

Learn how to rescue details, remove flyaways, add volume, and enhance the definition of hair in any photo. We break down every tool and technique in Photoshop to get picture-perfect hair, every time.

BV
Bart van der Wolf
Nov 1, 2003
"Bill Hilton" wrote in message
When I take large files from medium format film scans and shrink them to viewable jpeg or thumbnail sizes I lose a lot of the inherent image
quality.
This is taking files that might start out at say 11,000 x 8,800 pixels and reducing to 600 pixels wide for a web image or 90 pixels wide for a thumb.

I use one of 2 approaches.
1. Blur (radius depending on size reduction), and resample.
2. Filter|Pixelate|Mosaic…, single step averaging with Cell size 2=50%,
3=33.33%, 4=25%, etc., then resample with that percentage, and finish with the last bit of resizing to reach the required pixel size.

I’ve tried the one step approach and also have written actions to do it in
50%
increments and in 93% increments, which look better for most images. I’ve
also
tried re-scanning at 1,000 ppi instead of 4,000 ppi and this works well
too,
but I want to avoid scanning twice at all costs.
I have two questions …

1) Anyone have any other magic numbers besides 50% and 93% for something
like
this? Whipping up the action is pretty trivial so I can try different
values
but life is too short to try all of them :).

I haven’t found acceptable stair step settings. Resampling down, reduces edge contrast too much for my taste.

Bart
B
bhilton665
Nov 1, 2003
From: "Bart van der Wolf"

I use one of 2 approaches.
1. Blur (radius depending on size reduction), and resample.

Thanks Bart … I take it "blur" means Gaussian? Can you give me a starting point for radius, say for one of your 5400 dpi Minolta scans?

2. Filter|Pixelate|Mosaic…, single step averaging with Cell size 2=50%, 3=33.33%, 4=25%, etc., then resample with that percentage, and finish with the last bit of resizing to reach the required pixel size.

Will try this …

Bill
BV
Bart van der Wolf
Nov 1, 2003
"Bill Hilton" wrote in message
From: "Bart van der Wolf"

I use one of 2 approaches.
1. Blur (radius depending on size reduction), and resample.

Thanks Bart … I take it "blur" means Gaussian? Can you give me a
starting
point for radius, say for one of your 5400 dpi Minolta scans?

Yes, Gaussian is natural looking and it allows setting a radius. If you scale down by a factor of 2, I’d try a radius of 1 (so basically half the scale down factor) to start with.

2. Filter|Pixelate|Mosaic…, single step averaging with Cell size 2=50%, 3=33.33%, 4=25%, etc., then resample with that percentage, and finish
with
the last bit of resizing to reach the required pixel size.

Will try this …

And in method #2, which retains much more detail than the other methods, you can introduce a small radius blur (similar as in #1) for the final bit of resizing if you need to suppress aliasing.

If I e.g. reduce a 5400ppi 35mm scan to 5% of the original size with method #2, I still see roof tile detail, where the other methods return reddish mush.

As always after resampling, some small radius USM or high pass filtering will be needed.

Bart
NS
n8 skow
Nov 1, 2003
Why do you scan at such hi-res if your gonna resize everything down?

n8

When I take large files from medium format film scans and shrink them to viewable jpeg or thumbnail sizes I lose a lot of the inherent image
quality.
This is taking files that might start out at say 11,000 x 8,800 pixels and reducing to 600 pixels wide for a web image or 90 pixels wide for a thumb.
I’ve tried the one step approach and also have written actions to do it in
50%
increments and in 93% increments, which look better for most images. I’ve
also
tried re-scanning at 1,000 ppi instead of 4,000 ppi and this works well
too,
but I want to avoid scanning twice at all costs.

I have two questions …

1) Anyone have any other magic numbers besides 50% and 93% for something
like
this? Whipping up the action is pretty trivial so I can try different
values
but life is too short to try all of them :).

2) Anyone aware of a web site that has studied this in some detail in a systematic way? Got a link?

Thanks for any help.

Bill

B
bhilton665
Nov 1, 2003
From: "n8 skow"

Why do you scan at such hi-res if your gonna resize everything down?

Scanning to make large prints, I only put a few on the web so instead of "everything" it’s just a small % that need downsampling.

If my target was the web I’d probably scan at 1,000 dpi instead, but it isn’t so I don’t.

Bill
N
nospam
Nov 1, 2003
In article ,
(Bill Hilton) wrote:

From: "n8 skow"

Why do you scan at such hi-res if your gonna resize everything down?

Scanning to make large prints, I only put a few on the web so instead of "everything" it’s just a small % that need downsampling.
If my target was the web I’d probably scan at 1,000 dpi instead, but it isn’t so I don’t.

Why 1000spi? 120 will do!
NS
n8 skow
Nov 1, 2003
72dpi is defacto for web…

I’m not sure what your printing on, but still sounds like your wasting a lot of time. Offset press only handles 266-350 dpi (133-175 lpi)…

If on the other hand your talking bout printing to film, disregard my rambling.

=)
n8

If my target was the web I’d probably scan at 1,000 dpi instead, but it
isn’t
so I don’t.

Bill
BV
Bart van der Wolf
Nov 1, 2003
"jjs" wrote in message
In article ,
(Bill Hilton) wrote:
SNIP
If my target was the web I’d probably scan at 1,000 dpi instead, but it
isn’t
so I don’t.

Why 1000spi? 120 will do!

By scanning at a higher resolution, you can crop and resize by averaging or pixel binning. This will effectively reduce graininess, even of high speed film.

Bart
N
nospam
Nov 1, 2003
In article <3fa40151$0$58705$>, "Bart van der Wolf" wrote:

"jjs" wrote in message
In article ,
(Bill Hilton) wrote:
SNIP
If my target was the web I’d probably scan at 1,000 dpi instead, but it
isn’t
so I don’t.

Why 1000spi? 120 will do!

By scanning at a higher resolution, you can crop and resize by averaging or pixel binning. This will effectively reduce graininess, even of high speed film.

Good point. I did not consider that the scans might be of small images, possibly 35mm slides which are shown 4X or more. Regardless, I’d not mind comparing outcomes of Tri-X and D76 images done so. Might be interesting.

MacBook Pro 16” Mockups 🔥

– in 4 materials (clay versions included)

– 12 scenes

– 48 MacBook Pro 16″ mockups

– 6000 x 4500 px

Related Discussion Topics

Nice and short text about related topics in discussion sections