Digital photo editing advice

On Oct 12, 2004, at 9:25 AM, Kathleen Timms wrote:

> There is a lot of bad photography to fix, not to mention the years'
> accumulation of dust, cat hair, etc, so my questions for all you
> digital photography experts out there are:
> 1. should I be editing the full-size images first, and then resizing;
> or vice-versa?
> 2. does the order of Charles' editing procedure apply to my task?

#1 varies from image to image, but for the most part, you're probably
better off working with the reduced image. Fewer pixels means less
work. The exception to this would be if resizing causes a piece of dust
or a major problem in the photo to be magnified. You could handle that
on a photo-by-photo basis.

When you remove dust, you're more or less blurring the photo or
manually drawing over flaws in the picture (perhaps with a "healing" or
rubber stamp tool of some sort). In essence, you're destroying unwanted
pixels.

When you shrink an image, you're also destroying pixels.

What I might recommend trying, is to slightly blur your image first
(and possibly run a dust & scratches filter or do some of the major
easy repairs) and then shrink your picture. The blurring will be
reduced when it shrinks, and you can re-sharpen to bring focus back in.
Since you're viewing platform is a TV, super sharpness won't show up
anyhow (unless it's on high-definition), so this should save you a fair
bit of effort.

This would make the order of events somewhat different:

#1: Import
#2: Remove noise
#3: Dust & Scratches and/or blur
#4: Resize image
#5: Adjust color
#6: Remove remaining flaws in the image
#7: Crop & sharpen

Try it out, see what you think. Please post back to the list if this
doesn't work for you or if a modified approach works better. I'd love
to know. This is mostly hypothetical. :)

--Nik

On 12-Oct-04, at 11:25 AM, Kathleen Timms wrote:

> 1. should I be editing the full-size images first, and then resizing;
> or vice-versa?
> 2. does the order of Charles' editing procedure apply to my task?
>
> Playing around with the images in Photoshop has not demonstrated any
> definitive answer to question #1; but I do notice that using the
> Unsharp Mask brings out even more dust and dirt which wasn't
> noticeable before sharpening.
>
> Any advice for handling this massive project will be gratefully
> received.
>
> -=Kathleen (in rural Ontario)

Hi from urban Ontario. Just using Photoshop, I work in this order:

Gross colour adjustments
Levels or gamma curve adjustments
Fine (masked) colour adjustments if needed
Crop
Remove dust and scratches
Unsharp mask (or third party sharpening plugin)
Resize/save

If the colour is badly off, you cannot accurately adjust the tone. Once
those are done, you can see what you have and crop it. That saves you
from wasting time on areas you will discard. Always try to do all
editing on the full size image. Significant downsizing will reduce the
tendency for banding after severe level adjustments. Fix major dust and
scratches before sharpening. Worry about the tiny stuff only if it is
still objectionable after downsizing. Some of it will disappear.

If you are scanning slides, an anti-static lens cloth and either
compressed air or one of those squeeze bulb things from a photo supply
store are useful. There shouldn't be much dust in the images. If you
are paying for the scanning, there shouldn't be any. hth

robert mcgonegal

Posted by ktimms on 08:25am Oct 12, 2004:
>1. should I be editing the full-size images first, and then
>resizing; or vice-versa?

For the purposes of your DVD production it doesn't much matter, but
why not work on the full-size images? Then you'll be ready to
re-purpose them if the need ever arises. I know you said you didn't
want to print them, but what about High Definition TV? Will other
people need to use them, for other purposes? Perhaps you'll want to
change your mind about cropping at a later stage, particularly if you
have portrait format images.

All you have to lose is disk space, and I'd rate that cheap compared
to your time.

> ...prepare them for DVD slideshows, which I have learned through
>experimentation absolutely requires resizing the images from
>1840x1232 down to about 640x480 pixels

As you say, TV has a low resolution and consequently seldom does
justice to still photographic images. Since you've got higher
resolution scans available you could compensate with a bit of
"rotoscoping" - zooming and panning (dollying and tracking to use the
correct terms).

If you haven't got any fancy video editing software, you could do
this in iMovie - the Ken Burns Effect. [You don't need a video camera]

>2. does the order of Charles' editing procedure apply to my task?

Yes. Charles Maurer's procedure is general. It doesn't matter how you
intend to output your images. In this context printing your scans
onto paper is the same as "printing" them to DVD.

>...resizing the images from 1840x1232 down to about 640x480 pixels.
>I do this in two stages, saving each set of resizings in separate
>folders.

Have you automated the resizing? There are many ways to crunch whole
folder structures of images, but at times like this I reach for
GraphicConverter.

Andrew

Kathleen asks:

1. should I be editing the full-size images first, and then resizing or vice-versa?

FWIW my advice is always to do as much as you can with the full size image so long as you have enough CPU horsepower and don't have to wait while PS processes each task. The "full size" that you are describing is still only a 6.5 MB image (at 8 bits/channel). For elimianting defects with the clone tool, resolution always helps but I imagine you are doing more with the healing brush and patch tools where resolution is probably less important. And . . . if you don't have the latest version of PS where these tools are working at their best then your task would become a lot easier with an upgrade.

Another reason for working at the higher resolution is to keep a cleaned up version of each file least you ever want to repurpose the picture for any reason.

I know opinions on 'when to sharpen' vary but I try not to do any sharpening before I have an image at it's final size and output resolution. I'm therefore not surprized to hear that your dust and scratches are more pronouced if you have used the unsharp mask too early on in the process. Those objects have edges and that's what is accentuated with the USM filter.

FWIW I generally only use the USM filter on the lightness channel of an image when it is Lab mode. That way I can sharpen to my heart's content and never introduce any colour fringing.

Good luck with your mammoth task.

John Wolff email: dtopcompwave.co.nz Hamilton, NZ

At 8:25 AM -0700 10/12/04, Kathleen Timms wrote:

>My goal, for about 3500 digitally-scanned slides from the last 30
>years, is not to print them but to prepare them for DVD slideshows,
>...

I'm doing similar things to prepare for a DVD slideshow. What are
you using to make the slideshows? (iDVD, Toast, something else?)
Did you try other tools before selecting your choice?

>which I have learned through experimentation absolutely requires
>resizing the images from 1840x1232 down to about 640x480 pixels. I
>do this in two stages, saving each set of resizings in separate
>folders.

I resize as the second last step (but keep a copy of the larger
format), and then apply sharpening.

Dave

Those of you making photo-DVDs might want to try out Photo-to-Movie.

http://lqgraphics.com/software/phototomovie.php

Great little program and it seems to do a better job of making
attractive photo-DVDs than mucking around with the Ken Burns Effect.
There's a good review of it on Mac360, and it pretty well sums up
anything I could write here.

http://www.mac360.com/index.php/mac360/more/
the_absolute_best_mac_application_for_photos_and_video_period/

Note that it does nothing at all to help with cleaning up photos, but
it will help you prepare them and display them attractively on a DVD.
Another nice thing you can do is import the DV clips it produces into
iMovie and mix 'em up with some video clips, if you happen to have any.
I'm still kicking myself for not getting a videographer at my wedding
for just that reason.

--Nik

On Tuesday, October 12, 2004, at 11:25 AM, Kathleen Timms wrote:

> I have been studying Charles Maurer's amazingly informative article in
> TidBITS #748, and have summarized the steps he uses in the order he
> follows:
>
> <http://db.tidbits.com/getbits.acgi?tbart=07832>
>
> 1. import raw data from the camera
> 2. remove noise
> 3. remove blur
> 4. compensate for basic optical problems
> 5. adjust tonality and balance colour
> 6. remove dust and scratches
> 7. crop
> 8. sharpen
> 9. create a new file with the optimal number of pixels for the size
> of print and resolution of your printer
>
> My goal, for about 3500 digitally-scanned slides from the last 30
> years, is not to print them but to prepare them for DVD slideshows,
> which I have learned through experimentation absolutely requires
> resizing the images from 1840x1232 down to about 640x480 pixels. I do
> this in two stages, saving each set of resizings in separate folders.

Are you familiar with Photoshop's Actions and Droplets? These
automation features can make resizing and other routine tasks a snap
and save lots of time and aggravation:

http://www.digitalphotobook.net/tutorials/actions/actions_1.html

http://www.digitalphotobook.net/tutorials/actions/actions_9.html

>
> There is a lot of bad photography to fix, not to mention the years'
> accumulation of dust, cat hair, etc, so my questions for all you
> digital photography experts out there are:
> 1. should I be editing the full-size images first, and then resizing;
> or vice-versa?

You'll probably find that on many occasions you'll have to do some
major magnifying in order to see what you're doing, so I recommend
leaving the resizing till after you're done retouching.

> 2. does the order of Charles' editing procedure apply to my task?
>
> Playing around with the images in Photoshop has not demonstrated any
> definitive answer to question #1; but I do notice that using the
> Unsharp Mask brings out even more dust and dirt which wasn't
> noticeable before sharpening.

Charles was right to suggest that unsharp masking be the last
retouching task. Make sure all your layers are flattened first. I've
found it a good idea to "save as" first so that if you change your mind
about anything, or try something else, you've got the layers to work
from.

Did you try the healing brush before the unsharp mask? It's commonly
used to fix zits and wrinkles, but I've found scratches and dirt
respond well to the blending capabilities of this brush, esp. when you
use a tiny brush on a soft setting. There's good instructions on using
this brush in Photoshop Help.

Keep in mind that the unsharp mask's effects are usually more obvious
on screen than in print, so for a DVD, you should experiment with the
settings carefully.

Make sure you have the "Preview" selected so you can check the effect
of the mask before you apply it. You place the pointer in the image
preview and drag it around to check it out, and also use the plus and
minus buttons below the window to adjust the magnification.

Photoshop's "0" defaults for amount, radius and threshold will sharpen
all the pixels in the image, and unless you adjust the controls
appropriately, it's probably why the dust and dirt is more prominent -
it's increasing the contrast in the entire image.

I hope this helps,

Marilyn

Kathleen wrote:

>My goal, for about 3500 digitally-scanned slides from the last 30
>years, is ... to prepare them for DVD slideshows, ...
>resizing the images from 1840x1232 down to about 640x480 pixels.

NTSC DVD's store frames as 720*480 (non-square) pixels, so you would be underutilizing the medium to downsample the images so drastically. (And PAL DVD's have even more pixels.)

I haven't performed the following experiment, but I bet you'd see better results handing iMovie (or whatever) images that are 1280*960, letting the encoding software perform the downsampling.

>1. should I be editing the full-size images first, and then resizing;
>or vice-versa?

I can't claim much video experience, but I always edit full-size images, because that gives me more pixels to understand the image I'm cleaning up--even when highly magnified. Why guess whether a lone brown pixel on someone's downsampled face was a freckle or dust, when you could work with the original scan?

Doing pixel editing of a downsampled image is also harder because changing one pixel often requires altering an adjacent pixel in some nonobvious way, to avoid creating a new artifact worse than the one being corrected. I find this particularly problematic where teeth are concerned. It's better to edit the hi-res image, and let other software perform any antialiasing as a side-effect of downsampling.

In the case of video, there is another potential issue: convolution (an image transformation to minimize strobing on an interlaced television). You shouldn't need to worry about this, particularly if your target television handles progressive scan video, and your DVD authoring software knows how to encode slide shows in 480p (i.e. progressive) video, but this is one more reason I would give the encoding software more data to work with.

I have some questions:

If I drop a "high res" image into the Ken Burns effect in the middle of an iMovie project based on an interlaced MiniDV video, will the slide be degraded to interlaced quality, or is the whole project upsampled to progressive mode? (Or can DVD's jump back and forth?) Do iDVD slide shows look better than in-video slide shows? (They should save space, and potentially avoid certain MPEG artifacts, but do viewers see benefits?)

Would DVD Studio Pro render better images in a slide show?

--Christopher

Thanks to all of you who sent helpful tips.

On Oct 12, 2004, at 12:59 PM, Nik wrote:
> Please post back to the list if this doesn't work for you or if a
> modified approach works better. I'd love to know.

After hours and hours and hours! of experimenting, here's what I've
discovered works best. BTW, I am using OSX 10.3 and Photoshop Elements
2, not the full-size Photoshop. It has most of the editing features I
need, but not the scriptability.

1. Import the images as TIFFs. (ColorSync scripts don't work on PSD
files.)

2. Since we will be viewing the images on our TV, embed the NTSC colour
profile into them in order to see what they will look like. (Drag
folder onto re-written 'Embed profile' script). The colour differences
are amazing.

3. Edit each full size image at 100% or 'actual pixels' view. This
seems to be important to PSE for the following edits:
        Remove noise first. This one tip from Charles Maurer makes all the
difference to the end result. In PSE 2, use the Noise/Despeckle filter.
        Adjust the Levels. Sometimes PSE's Auto Levels looks best, sometimes I
adjust the sliders manually.

4. Crop. Looking at the image at 100% in step 3 often shows me how to
best improve the composition. Most of these old photos need improving!

5. Save AS same filename to another folder. (With 3500 slides to edit,
I need some way to keep track of which ones I've done.)
        For the second editing session:

6. Remove dust & scratches. For this task, I have to select the areas
in the photo with blobs of dust or hair, then use the
Noise/Dust&Scratches filter. Using the filter on the whole photo just
produces a blurry mess, but using the filter (with a radius of 2-3
pixels and a threshold of 12-16) on selected areas gives a very nice
result. This is the most time-consuming portion of the whole process.

7. Correct the colour, if necessary.

8. Remove red-eye, if necessary, and/or correct any other flaws.
        Save the full-size, edited TIFF file for possible, other future uses.

9. When all the files in that specific folder have been edited, batch
process the whole folder to resize each image to 960 pixels wide.
Vertical photos have to be rotated sideways before doing this step. The
resized photos are saved in a separate file.

10. The resized photos are then resized again, to 480 pixels high.
Vertical photos have to be rotated right-side-up before doing this
step. This second resizing results in a 717x480 horizontal image, which
is pretty close to the 720x480 NTSC DVD standard.

11. Sharpen the twice-resized images, using the Unsharp Mask. Then
batch process them all into JPEGs for use in an iDVD slideshow.

> #1 varies from image to image, but for the most part, you're probably
> better off working with the reduced image. Fewer pixels means less
> work.

        Actually, my experiments show that working on full-size images for
steps 1-8 gives better results. This might be a Photoshop Elements
quirk, or might be standard for working on digital images in general.

On Oct 18, 2004, at 5:09 AM, ChristopherSchmidtZone.com wrote:
> I haven't performed the following experiment, but I bet you'd see
> better results handing iMovie (or whatever) images that are 1280*960,
> letting the encoding software perform the downsampling.

        I have performed the experiments (wasting a DVD in the process) and
believe me, iPhoto, iMovie and iDVD do a really bad job of
downsampling. Which is why I decided to use Photoshop Elements for the
whole process.

I hope this outline might be useful to anyone else attempting the same
task, and once again, thanks to everyone who replied to my question,
and especially thanks to Charles Maurer for his motivating article.

-=Kathleen (staring at her computer in rural Ontario)

>2. Since we will be viewing the images on our TV, embed the NTSC colour
>profile into them in order to see what they will look like. (Drag
>folder onto re-written 'Embed profile' script). The colour differences
>are amazing.

Shouldn't you be using a (re-written) "Match to specific profile" script instead of "Embed profile"? Unless your scanner driver is converting the color space to NTSC for you, it makes no sense to embed the NTSC profile into a scanned file. (If your scanner driver does convert the color space, it probably adds the appropriate profile to the output already.)

Personally, I set up Photoshop to ask me what to do about color matching on each untagged file I open.

> Adjust the Levels. Sometimes PSE's Auto Levels looks best, sometimes I
>adjust the sliders manually.

I bet you're compensating for the inappropriate embedded profile in this step.

>...resize each image to 960 pixels wide...
>
>10. The resized photos are then resized again, to 480 pixels high.
>... This second resizing results in a 717x480 horizontal image, which
>is pretty close to the 720x480 NTSC DVD standard.

Hmmm. NTSC DVD pixels aren't square, so I don't think this should work as hoped. I bet you'll find your "wide" images are being scaled down a little and letterboxed. Photoshop CS does have a pixel-aspect-ratio command in the Image menu to indicate non-square pixels, but it doesn't look as though iDVD (or any other program I tried) respects it. I'm not convinced Photoshop was tagging saved files using their native representations of non-square pixel aspect ratios, but I didn't investigate thoroughly.

Worse still, it appears that iDVD scales images down into the title-safe region of the video frame, so even more downsampling is at work.

>On Oct 18, 2004, at 5:09 AM, ChristopherSchmidtZone.com wrote:
>> I haven't performed the following experiment, but I bet you'd see
>> better results handing iMovie (or whatever) images that are 1280*960,
>> letting the encoding software perform the downsampling.
>
> I have performed the experiments (wasting a DVD in the process) and
>believe me, iPhoto, iMovie and iDVD do a really bad job of
>downsampling. Which is why I decided to use Photoshop Elements for the
>whole process.

Fascinating. I guess my input was misleading all around. Sorry!

Incidentally, you can burn to DVD-RW media by erasing a disc in Disk Utility, and selecting iDVD when the Finder asks you what to do with the blank disc.

Thanks to Nik Gerber's post earlier in this thread, I tried out PhotoToMovie just now, and I am pleased to report that it fills the entire 720 * 480 DVD frame with my images, appropriately scaled in each dimension. At this point it's what I plan to use when I make a (long overdue!) slide show DVD.

    <http://lqgraphics.com/software/phototomovie.php>

PhotoToMovie has export options for DV or "full DV", which it says are for iMovie and iDVD respectively. Whether "full DV" produces better output on progressive players, I cannot say, but it's one more thing to try.

IMO PhotoToMovie is worth your time to investigate (but take Nik's word, not mine!).

--Christopher