Quick De-cracking Action results

[DJ Dubovsky]

OK Tim, here's the results I got running the De-cracking action on 3 different cracked photos. I also compared the various High pass settings to see the results.

Here's the original version of the the first photo. It's the family photo from the Beyond Redemption restoration challenge.
DJ

[DJ Dubovsky]

Here's the same photo using the minimum high pass setting of 0.4. You can see a definate difference using the de-crack technique.
DJ

[DJ Dubovsky]

Here's the same photo again at the medium setting of 1.5. The difference although significant from the original bares little difference from the minimum setting.
DJ

[DJ Dubovsky]

Here's that photo at the high setting of 4.0. It seems to be little difference from the lowest setting of 0.4. I still think this is a good technique but I thought these results were interesting.
DJ

[DJ Dubovsky]

Here is the second photo original condition.
DJ

[DJ Dubovsky]

Here is that same photo after running de-crack on it. This was done at a high pass setting of 5.0 which is higher then the heaviest setting. There is little change in this photo. Not the drastic difference I noticed in the family photo.
DJ

[DJ Dubovsky]

Here's the third photo I tried it on. This is the original photo.
DJ

[DJ Dubovsky]

Here it that same photo at a high pass setting of 5.0. Again there is not much difference from the original. I am not sure what the difference is from the first photo or other photos that it worked significantly on. This is just a demonstration of the varied results that can be expected in using this technique.
DJ

[Doug Nelson]

Wow, DJ. That was a lot of work, and some very interesting results. It appears that this techique works best when the substrate paper is showing through (ie: much whiter than anything in the image). I guess we could cheat and manipulate the curve beforehand to make sure that was the case, then adjust back afterwards.

I notice in the bar scene he looks a little sleepier in the actioned one than in the original. Some eye detail changed. That's what makes me think this should be done only after we make sure the paper is the whitest thing in the image.

I wonder if there's an equal and opposite action for dark cracks and spots. Tim?

[DJ Dubovsky]

Thanks Doug. It really wasn't that much work with that action. If I had to do it manually that might be the case.

Interesting observations on the bar scene and your ideas for building upon it's abilities.
Dj

[Ed_L]

Hey Debbie,

You really put it through it's paces. Interesting results for sure.

Ed

[Tim Edwards]

oops! Missed this thread!

Debbie,

Sorry, I just noticed the thread... I'll take a deeper look, but some initial thoughts:

The Decrack action's intent is to isolate high-contrast noise and fill it with the 'minimum' filter, leaving the rest of the image alone if possible. The setting does not affect how quickly large cracks are filled, it simply determines how large and contrasty edges must be to be found. So, if you run the Decrack action at 0.4 for 3 iterations, you'll get the same amount of fill in large cracks as if you ran it at 1.5 the same number of times, but you'll have less impact on the actual image (assuming it's of low enough contrast).

Another limitation is that the filter uses the minimum filter to fill damage, meaning that it will grab dark values from nearby. Thus, dark damage will be amplified...

The first image looks like it has both the light/dark damage problem as well as having high-contrast elements in the image itself...

_Tim

[DJ Dubovsky]

Tim
I ran each version only once. I know the large cracks wouldn't be affected much and like I said, I love what it does when it works. My curiosity was on the first photo where I ran all 3 levels, one on each copy from the weakest to the heaviest setting but for sorme reason they all seem the same. I would have figured the difference would have been more noticable. Don't mistake my little observation as a put down. You did say it worked best under certain circumstances and I am thouroughly impressed with the results when it does work. It's still a valuable tool and one I shall use often, I'm sure. I was just curious and figured you would know what had happened in these situations. Thanks for responding Tim. I know sometimes these threads hide and I don't know why.
DJ

[Tim Edwards]

an animation

Debbie,

Here's an animation of the two extreme settings, 0.4 and 4.0, on your first example. I think the differences will be clear to you... They've been highlighted with a hue shift. Note that the 0.4 setting affects less of the actual image.

Sorry, the full-sized version was too big to upload, so I cropped and downshifted.

I don't mind the questions or observations... Maybe I can actually improve the thing!

_Tim

[DJ Dubovsky]

OH Wow, I can see the difference now. It was hard to see the difference on my versions although I knew it should be there. That really brings it out well. Thanks Tim.
DJ