What’s your “image rehab” routine?

GrannyGrump · 10-25-2013, 12:46 AM

I need some faster procedures for processing illustrations. I am in a sequence of heavily illustrated books, numbering in the hundreds for each book. I find that I am spending way too much time cleaning these manually.

I do use a lot of scripts in my image editor (PaintShop Pro v.9). I can batch process with my scripts -- for the first stage of initial cleaning. But due to the non-uniform results of paper aging, dirt, and damage, I still spend a lot of time manually using clone brush or smudge or dodge/burn, etc etc. I find that if I use scripts *too* much for cleaning, they bleach out way too much detail.

Does anybody have some time-saving tips?

Jellby · 10-25-2013, 03:25 AM

Not really a time saver, but it might be if you are not so much a nitpicker as I am. I use this for black on white illustrations, with mainly white background:

- Open original raw scan (from archive.org, or from my own scan)
- Convert to grayscale, straighten (measure angle and rotate arbitrary), crop
- Add a completely black layer on top, but set the blend mode to the one that makes everything black except pure white pixels in the background. In GIMP this is called "Burn".
- Add a transparent layer between the background and the black one.
- Open the "Levels" dialog for the background layer, move the controls so what should be white shows through the top black layer, and the lower peak in the histogram is below the black level.
- Cover the speckles in the background by painting white on the middle transparent layer (this makes it easy to delete if I paint too much, better than undo). This is the most time-consuming part, but it really does make a difference (to me).
- When satisfied, disable the top black layer, flatten, save as PNG for storing, scale down if desired and save as JPG for using.

GrannyGrump · 10-25-2013, 04:31 AM

Jellby, thank you for the tip about the "burn" overlay together with "levels". This could work out well for me. I have not worked much at all with Levels or Histogram. Must-conquer-fear, fear is the mind-killer.

I have been using a PSP dialog for "HSL" (Highlight, Midtone, Shadow) adjustment, where you use positive and negative settings to increase the strength of the - you guessed it - highlights, midtone, shadows. It works ok, but there is a lot of overlap between a dark-colored highlight and a light-colored midtone. And then using paintbrush and dodge-brush to bleach out lingering specks, but yes, that is tedious. Your way sounds simpler.

Thanks again. (I really need to try GIMP again, maybe I can learn to love it.

)

SBT · 10-26-2013, 12:42 PM

I use imagemagick for batch processing. Sample from Unix shell, with typical numbers for b&w photos & drawings. Numbers vary depending on image exposure, dimensions etc.

Code:

mkdir clean
for x in *.jpg
do
convert $x -colorspace gray \ #convert to B&W
-level 10%,75%,0.6 \ # adjust B/W levels
-unsharpen 5x1+0.4+0.02 \ # makes lines less fuzzy
-fuzz 20% -trim \ # autocrop, but I seldom use it
-resize 25% \# shrink
-colors 16 \# ...since e-ink only has 16 shades anyway...
clean/${x%jpg}png \ # convert to png
done

... and then run them through pngcrush afterwards to save another few bytes.

Jellby · 10-26-2013, 01:06 PM

Quote:

Originally Posted by SBT

Code:

-colors 16 \# ...since e-ink only has 16 shades anyway...

Hmm... but are the 16 colors chosen by ImageMagick the same as the 16 shades supported by the reader?

Some years ago, e-ink supported only 4 grey levels... Will you upgrade all your pictures when a newer device comes up? What if I want to read it in a color tablet with 256 levels?

SBT · 10-26-2013, 04:03 PM

Well, future-proofing stuff is tricky...
You can match the 16 shades in imagemagick to exactly match e-ink shades, but I've never bothered – worst display approximation on the display will be within 3% of the original.
For b&w line drawings, I find 16 shades sufficient even on computer screens.
I like to keep the epub files below 20MB, so I have to squeeze the image quality somehow on the most thoroughly illustrated books. I can choose between reducing resolution, quality, or colour space. I go for colour space, and keep slightly higher resolution as a nod to futureproofing.

Freeshadow · 10-27-2013, 08:31 PM

@ SBT why pngcrush when pngout is better?

SBT · 10-28-2013, 03:12 AM

Quote:

Originally Posted by Freeshadow

@ SBT why pngcrush when pngout is better?

Mental inertia...

Jellby · 10-28-2013, 04:01 AM

Quote:

Originally Posted by Freeshadow

@ SBT why pngcrush when pngout is better?

pngcrush is open source, pngout is not. Other than philosophical/ethical concerns, that means pngcrush will be more easily available for Linux users (use the distribution installation channels, rather than search-download-store the Windows way).

Freeshadow · 10-28-2013, 04:07 AM

The console version of pngout is free & available for Linux and mac too. Only the win-GUI and Photoshop plugin versions are to be paid for.

Jellby · 10-28-2013, 05:04 AM

Quote:

Originally Posted by Freeshadow

The console version of pngout is free & available for Linux and mac too. Only the win-GUI and Photoshop plugin versions are to be paid for.

I never said it was not free, I was talking about "open source". And I didn't say it was not available for Linux, but not so convenient.

Freeshadow · 10-28-2013, 06:43 AM

True. A FOSS purist would avoid it.

derangedhermit · 11-10-2013, 10:51 AM

I'd like more threads on preparing images for use in ebooks. I have passed many a pleasant day in rehabbing old photos and artwork to my satisfaction (or as close as I could get). And that's a problem for productivity, isn't it - that it takes so long?

I guess the guidelines stay the same no matter the image, but I'd think the specific routines and/or parameters would differ for different image types - e.g. line art v. continuous tone. Is there a tutorial?

It would also be nice to hear how people deal with the fleurons, dingbats, dividers, etc that are commonly encountered in scans and often carried over to ebooks. It seems to me to use images of dividers is bad form. It certainly seems a waste when there are many images of the same symbol, one clipped from the scan at each place encountered. One could do better using transparency (if it is well supported), but it seems best to simply do something else (in most cases, I'm sure the case can be made for exceptions).

This thread reminded me - I'm off to the MR library to see if anyone has attempted William Blake using full-page original illustrations alone.

SBT · 11-10-2013, 01:55 PM

I've discovered that you're not home and dry with png images even if you do use 8-bit colour space. For my Norwegian editions of "Three in Norway – by two of them" I used the images from PG, but reduced the number of colours to 16. The original language edition uses png, the modernized one uses exactly the same images converted to gif. The png versions show up all fuzzy on my PRS-T1, the gif ones are all right.

Darn....

SBT · 11-10-2013, 02:09 PM

@derangedhermit:
I posted a mini-tutorial (tutorialette?) on how to use Gimp for basic image rehab some time ago.
Personally, I use imagemagick and batch processing whenever there are many images and I can get away with it.
Regarding fleurons and such, I seldom come across any which it is worth transferring to the ebook. However, in those cases, I normally spend some time converting them as neatly as possible to svg. The only book I can remember I have done it in is English as she is spoke
Otherwise, I use stock fleurons from a free fleuron font. However, I normally convert them to svg, rather than importing the font. It's more robust, scales better, and degrades better that way.

10-25-2013, 12:46 AM	#1
GrannyGrump Obsessively Dedicated... Posts: 3,200 Karma: 34977896 Join Date: May 2011 Location: JAPAN (US expatriate) Device: Sony PRS-T2, ADE on PC	What’s your “image rehab” routine? I need some faster procedures for processing illustrations. I am in a sequence of heavily illustrated books, numbering in the hundreds for each book. I find that I am spending way too much time cleaning these manually. I do use a lot of scripts in my image editor (PaintShop Pro v.9). I can batch process with my scripts -- for the first stage of initial cleaning. But due to the non-uniform results of paper aging, dirt, and damage, I still spend a lot of time manually using clone brush or smudge or dodge/burn, etc etc. I find that if I use scripts too much for cleaning, they bleach out way too much detail. Does anybody have some time-saving tips?

10-26-2013, 12:42 PM	#4
SBT Fanatic Posts: 580 Karma: 810184 Join Date: Sep 2010 Location: Norway Device: prs-t1, tablet, Nook Simple, assorted kindles, iPad	I use imagemagick for batch processing. Sample from Unix shell, with typical numbers for b&w photos & drawings. Numbers vary depending on image exposure, dimensions etc. Code: mkdir clean for x in *.jpg do convert $x -colorspace gray \ #convert to B&W -level 10%,75%,0.6 \ # adjust B/W levels -unsharpen 5x1+0.4+0.02 \ # makes lines less fuzzy -fuzz 20% -trim \ # autocrop, but I seldom use it -resize 25% \# shrink -colors 16 \# ...since e-ink only has 16 shades anyway... clean/${x%jpg}png \ # convert to png done ... and then run them through pngcrush afterwards to save another few bytes.

10-26-2013, 04:03 PM	#6
SBT Fanatic Posts: 580 Karma: 810184 Join Date: Sep 2010 Location: Norway Device: prs-t1, tablet, Nook Simple, assorted kindles, iPad	Well, future-proofing stuff is tricky... You can match the 16 shades in imagemagick to exactly match e-ink shades, but I've never bothered – worst display approximation on the display will be within 3% of the original. For b&w line drawings, I find 16 shades sufficient even on computer screens. I like to keep the epub files below 20MB, so I have to squeeze the image quality somehow on the most thoroughly illustrated books. I can choose between reducing resolution, quality, or colour space. I go for colour space, and keep slightly higher resolution as a nod to futureproofing. Last edited by SBT; 10-26-2013 at 04:11 PM.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Image on click change to nex image for Ipad view.	vinayaksanga	ePub	9	04-29-2013 06:49 AM
inspect image properties /replace image	cybmole	Sigil	6	02-05-2013 12:46 PM
Different Send to Device routine based on the library.	iatheia	Library Management	10	01-24-2012 12:06 PM

10-25-2013, 03:25 AM	#2
Jellby frumious Bandersnatch Posts: 7,516 Karma: 18512745 Join Date: Jan 2008 Location: Spaniard in Sweden Device: Cybook Orizon, Kobo Aura	Not really a time saver, but it might be if you are not so much a nitpicker as I am. I use this for black on white illustrations, with mainly white background: - Open original raw scan (from archive.org, or from my own scan) - Convert to grayscale, straighten (measure angle and rotate arbitrary), crop - Add a completely black layer on top, but set the blend mode to the one that makes everything black except pure white pixels in the background. In GIMP this is called "Burn". - Add a transparent layer between the background and the black one. - Open the "Levels" dialog for the background layer, move the controls so what should be white shows through the top black layer, and the lower peak in the histogram is below the black level. - Cover the speckles in the background by painting white on the middle transparent layer (this makes it easy to delete if I paint too much, better than undo). This is the most time-consuming part, but it really does make a difference (to me). - When satisfied, disable the top black layer, flatten, save as PNG for storing, scale down if desired and save as JPG for using.

10-25-2013, 04:31 AM	#3
GrannyGrump Obsessively Dedicated... Posts: 3,200 Karma: 34977896 Join Date: May 2011 Location: JAPAN (US expatriate) Device: Sony PRS-T2, ADE on PC	Jellby, thank you for the tip about the "burn" overlay together with "levels". This could work out well for me. I have not worked much at all with Levels or Histogram. Must-conquer-fear, fear is the mind-killer. I have been using a PSP dialog for "HSL" (Highlight, Midtone, Shadow) adjustment, where you use positive and negative settings to increase the strength of the - you guessed it - highlights, midtone, shadows. It works ok, but there is a lot of overlap between a dark-colored highlight and a light-colored midtone. And then using paintbrush and dodge-brush to bleach out lingering specks, but yes, that is tedious. Your way sounds simpler. Thanks again. (I really need to try GIMP again, maybe I can learn to love it. )

10-27-2013, 08:31 PM	#7
Freeshadow temp. out of service Posts: 2,792 Karma: 24285242 Join Date: May 2010 Location: Duisburg (DE) Device: PB 623	@ SBT why pngcrush when pngout is better?

10-28-2013, 04:07 AM	#10
Freeshadow temp. out of service Posts: 2,792 Karma: 24285242 Join Date: May 2010 Location: Duisburg (DE) Device: PB 623	The console version of pngout is free & available for Linux and mac too. Only the win-GUI and Photoshop plugin versions are to be paid for.

10-28-2013, 06:43 AM	#12
Freeshadow temp. out of service Posts: 2,792 Karma: 24285242 Join Date: May 2010 Location: Duisburg (DE) Device: PB 623	True. A FOSS purist would avoid it.

11-10-2013, 10:51 AM	#13
derangedhermit Addict Posts: 239 Karma: 1280000 Join Date: Oct 2010 Location: USA Device: None	I'd like more threads on preparing images for use in ebooks. I have passed many a pleasant day in rehabbing old photos and artwork to my satisfaction (or as close as I could get). And that's a problem for productivity, isn't it - that it takes so long? I guess the guidelines stay the same no matter the image, but I'd think the specific routines and/or parameters would differ for different image types - e.g. line art v. continuous tone. Is there a tutorial? It would also be nice to hear how people deal with the fleurons, dingbats, dividers, etc that are commonly encountered in scans and often carried over to ebooks. It seems to me to use images of dividers is bad form. It certainly seems a waste when there are many images of the same symbol, one clipped from the scan at each place encountered. One could do better using transparency (if it is well supported), but it seems best to simply do something else (in most cases, I'm sure the case can be made for exceptions). This thread reminded me - I'm off to the MR library to see if anyone has attempted William Blake using full-page original illustrations alone.

11-10-2013, 01:55 PM	#14
SBT Fanatic Posts: 580 Karma: 810184 Join Date: Sep 2010 Location: Norway Device: prs-t1, tablet, Nook Simple, assorted kindles, iPad	I've discovered that you're not home and dry with png images even if you do use 8-bit colour space. For my Norwegian editions of "Three in Norway – by two of them" I used the images from PG, but reduced the number of colours to 16. The original language edition uses png, the modernized one uses exactly the same images converted to gif. The png versions show up all fuzzy on my PRS-T1, the gif ones are all right. Darn....

11-10-2013, 02:09 PM	#15
SBT Fanatic Posts: 580 Karma: 810184 Join Date: Sep 2010 Location: Norway Device: prs-t1, tablet, Nook Simple, assorted kindles, iPad	@derangedhermit: I posted a mini-tutorial (tutorialette?) on how to use Gimp for basic image rehab some time ago. Personally, I use imagemagick and batch processing whenever there are many images and I can get away with it. Regarding fleurons and such, I seldom come across any which it is worth transferring to the ebook. However, in those cases, I normally spend some time converting them as neatly as possible to svg. The only book I can remember I have done it in is English as she is spoke Otherwise, I use stock fleurons from a free fleuron font. However, I normally convert them to svg, rather than importing the font. It's more robust, scales better, and degrades better that way.