Detecting bad photos in camera and after


As I've noted, with digital cameras we all take tons of photos, and the next task is to isolate out the winners. I've outlined better workflow for this and there are still more improvements we need in photo management software, but one task both cameras and photo management software could make easier is eliminating the plain bad shots.

I've always wanted the camera to have a display mode that immediately shows, at 1:1, the most contrasty (sharpest) section of a photo I have taken. If I look at that, and see it's blurry then I know the whole photo is blurry, whether it be from camera shake or bad focus. If it's sharp but not the thing I wanted to emphasize, I may realize the autofocus found the wrong thing. (My newest camera shows in the review pane what autofocus points it used, which is handy.)

Indeed, if a camera finds that there is no section of the photo which is sharp, it might even display or sound a warning. Yes, sometimes I will take shots of fuzzy clouds where this will be normal. I can handle the false warning then. It might be so dark I can't get a good shot and will also ignore the warning, but other times it might tell me to shoot that one again.

(Nikon cameras have a feature where they take 3 shots and keep the sharpest of them. That's handy, but I still want to know if the sharpest of them is still no good.)

The camera could go further. With more sensitive accelerometers, it could actually calculate how much the camera rotated while the shutter was open, and since it also knows the focal length, it could calculate the amount of motion blur there will be in the shot. Again, it could warn you when it's too much, and tag this acceleration data in the EXIF fields of the file. Yes, sometimes one takes a tracking shot where you pan on a moving object and deliberately blur the background. In theory the detection of sharp objects in the field would reveal this, but in any event you can also just ignore the warning here.

For those will full flash cards, such detection could help in removing turkeys when you have to delete.

Until our cameras can do this, our photo management software could help. As noted, the first task in photo management is to divide the photos into groups. I divide into 5 groups myself -- bad shots, boring shots, average shots, winners and super-winners. Winners go into the slideshow for the particular shooting trip, super winners will go into a "best of the year" category.

The photo management software could scan over the photos, and find ones that are blurry. It could then let me do a quick scan over them, either as large thumbnails, or perhaps again showing me at 1:1 zoom the highest contrast crop. I could quickly pull out any pictures I still want and relegate the others to the bad photo pile, or even delete them. The same could apply for images that are obviously overexposed and underexposed. Again, I will still scan to see if there is anything to save, and in the case of the underexposed, I can do the scan in a mode where a compensation is done to brighten them to see what can be recovered. But after that, I don't want them in the way of my real workflow, to find the winners.


Not really related to your article, but just a general
question. Having finally purchased a digital camera,
I am thinking about what data to save. The camera
can save JPEG and raw data simultaneously. I'm thinking
of saving high-quality JPEGs and raw data for all shots.
The JPEGs will probably be good enough for printing
things out, even enlargements, but if I really want
to go back and reprocess the raw data, I could.

Basically, with just JPEGs, what you see is all you've
got. With just raw, all the images have to be processed
before getting a useful, practical format. With both,
I have the best of both worlds, at the expense of saving
the raw data (about 20 MB a shot). I can still get more
than a hundred combination shots on a memory card, though,
and don't take enough pictures that disk space will be
a problem. (I take about 100-200 per year, or at least
have in the past. I'll probably take a lot more with the
digital camera, but take several shots where I took one
before then keep the best one and delete the others from
the memory card.)

What do most people do?

Raw is of course the most complete saving of data, but a pain to work with, and provides minimal benefit for 98% of your shots. Where it can provide benefit include shots where you might want to change the colour balance, or night shots and shots where you wish to alter exposure or bring out more shadow etc.

Many photographers don't want to give that up and shoot raw. But the disk space/flash cost is quite high -- raw + large superfine jpeg means 4-5 times the disk space.

Now disk space is cheap and getting cheaper, so more often it's about the flash card space. But those are also getting cheaper. So most folks will probably go your way in time.

A slight saving is to shoot raw + a smaller jpeg, but it is only a slight saving. The idea being that once you have identified your "keeper" images you then go back to the raw files to work from them. It would be nice if tools made that easier to do -- catalog, tag and play around with small jpegs, then do a batch command to pull out the raws of those same images.

Jpeg compression artifacts are not a big issue on superfine jpegs. It's the loss of bit depth and a bit of colour info that are key.

"I’ve always wanted the camera to have a display mode that immediately shows, at 1:1, the most contrasty (sharpest) section of a photo I have taken."

What a weird way to determine if a photo is good. Sure, sharpness is important, but hardly the end all, be all of good photography. Many of the most famous, most revered photos in history are not sharp. Having your nose buried in your photos at an individual pixel level is just plain idiotic.

Some photos you want the subject sharp. Just because sometimes you may not have this as a priority doesn't mean it isn't a common goal. And when I take a shot in which I wanted the subject sharp, I want to know if it isn't. That's what showing me the sharpest parts of the photo would tell me, and tell me quickly.

I liked the accelerometer idea. Feels like idiot proofing, but it's nice. But automation is so hard here.

For instance, the idea of photo management software scanning over the photos, and find ones that are blurry... Some of our best shots ever are wide-aperture images with no focal depth - most of the shot is out of focus, and that's the point.

Worse yet, the idea of suggesting shots to delete based on poor optical properties is not going to be loved if it suggests the one and only last shot of your .

I think most of these problems need an aesthetic mind, and the cameras are not up to that task.

Here's one idea:
1. Camera space is cheap - The camera could also take and store bracket shots, deleting them if it runs low on space for new shots. Brackets could include exposure and auto focus modes. If power consumption was low enough and light high enough, it could store maybe 20 images - compressing them as diffs.

Then in iPhoto or whatever, when you don't like a main shot, the software can show you a cloud of bracket shots, perhaps offering to stitch them or use them to rescue over and under exposed regions into one good shot. And if you do like the shot, you can explore pro options for post-hoc varying the aperture, exposure etc.

Marketing could say “Take those ‘pro’ manual shots with the new GE auto!”

While bracketing is useful, one of my goals here is to speed up what has become one of the major time components of photography -- selecting shots. I'm not saying to delete the only picture of grandma, but to make it easier to pull out the best pictures of her. If the system says "none of those shots were any good" then the user can take the time to hand pick with the human eye.

Add new comment