How to convert scans to reasonable size?

[svga]

First my environment:
Hardware: Epson Perfection V200 Photo scanner
Software: xsane-0.995 (Linux)

I scan magazines at 300dpi with this environment, which produces PNM files which are about 25MB (2544x3509 pixel, depth 24). Of course I can't (or am not willing) upload these images directly to an image hoster. Therefore I currently convert the images with the convert program from ImageMagick:

convert -resize '1680x1680>' infile outfile.jpg;

This produces images with a maximum side length of 1680 pixels, the image size is about 300 KB.

My question: Is there a better way to compress the images to a reasonable size?

[Xxphd]

I always batch-convert with GraphicConverter

PICT 32 bit -> jpeg

[svga]

Thanks for the response, butI have to rephrase my question: Is there a better way to get a better quality/image size (in bytes) ratio?

One example: I've converted the same scan with two different options to approximately the same size (380 kB):

1st Conversion: convert -resize '1680x1680>' page26.pnm page26-1680x1680.jpg (uses standard compression 85):


2nd Conversion: convert -quality 27 page26.pnm page26-q27.jpg (this sets the "JPEG compression" to 27 without changing the image size):


Which one is better?

Of course proposals for other programs which produce better results are welcome

[icu]

Panorama : resize / resample



left origin 95%, middle resize + zoom , right resample + zoom. ( irfanview )

The resample function in Irfanview softens during the resample and reduces the filesize.



Panorama: jpg compression



from left to right:
100%, 501KB
60%, 69KB
95%, 302KB
85%, 156KB

It can be useful to save with high compression values ( 85%+ ) if the pic contains reflexions ( metal, fingernails, hair, ... ). Lower compression values decreases the depth effect due to lower brilliance.

[Jism Jim]

Originally Posted by svga

I scan magazines at 300dpi with this environment, which produces PNM files which are about 25MB (2544x3509 pixel, depth 24).

Can your scanning software produce bitmap files too? If so, this will increase your choice of converting software.
I don't know about Linux, but I use Irfanview myself. IMO it produces very good results with the quality setting at 80%.

[svga]

Originally Posted by Dario Rosselini

Can your scanning software produce bitmap files too? If so, this will increase your choice of converting software.
I don't know about Linux, but I use Irfanview myself. IMO it produces very good results with the quality setting at 80%.

No, the software itself can only produce PNM, but the formats are similar and I can convert to BMP with another tool without any loss of information.

[Xxphd]

Originally Posted by svga

I scan magazines at 300dpi with this environment, which produces PNM files which are about 25MB (2544x3509 pixel, depth 24).

producing a BMP or PICT will reduce the size to about 50% or even less of that I guess

[svga]

Originally Posted by Xxphd

producing a BMP or PICT will reduce the size to about 50% or even less of that I guess

No, it's almost the same size (26781040 respective 26780742 bytes in the example):

> convert page26.pnm page26.bmp # converts PNM to BMP (lossless)
> ls -l
-rw-r-----  1 anon anon 26781040 Jun 26 23:43 page26.pnm
-rw-r--r--  1 anon anon 26780742 Aug  8 18:41 page26.bmp

As far as I know PNM has also an ASCII representation instead of using raw binary data. If this would be used you would be right.

PNM and BMP are almost the same formats, they store for every pixel the full information (e..g. 24 bits RGB value or 8 bit grey value).

[tuffy]

Originally Posted by svga

Thanks for the response, butI have to rephrase my question: Is there a better way to get a better quality/image size (in bytes) ratio?

One example: I've converted the same scan with two different options to approximately the same size (380 kB):

1st Conversion: convert -resize '1680x1680>' page26.pnm page26-1680x1680.jpg (uses standard compression 85):


2nd Conversion: convert -quality 27 page26.pnm page26-q27.jpg (this sets the "JPEG compression" to 27 without changing the image size):


Which one is better?

Of course proposals for other programs which produce better results are welcome

Which one is better?

The first picture.

Why?

Because the second picture has too many JPEG artifacts.(the Quality=27 is too low)

There is a problem here, the first picture's size is 1218x1680 (resized 50%), the second picture's size is 2544x3509. It's not easy to compare them when they have different sizes.

Is there a better way to get a better quality/image size (in bytes) ratio?

No. If you want good quality you pay with a bigger file size.

The JPEG compression is lossy, you can use it with Quality setting between 70-100, 85 is a good compromise.

Some HQ (big sizes) examples:

1. 200dpi descreen on - 600kB


3. 300dpi descreen off, edited - 1MB


5. 300dpi descreen on, edited - 1MB


It's interesting that the 300dpi pic with descreen ON has a little moire, and with descreen OFF no moire.

Tuffy

[beutelwolf]

Back in the days when I was fiddling with these options, my rule of thumb was: scan with high resolution, reduce afterwards - although I left the reduction to a picture editor (PaintShop Pro), not the scanning software. Though this may have also been influenced by the tiny little issue that my first scanners didn't support picture reduction.

Anyway, that strategy that seemed to help with Moire patterns. When I got Moire with scanning at 300dpi I moved to 500pdi or even 1000dpi, creating a monster-sized file, and then let the picture editor do its magic afterwards.