So all the scans I do at minimum 2000 dpi are over the top? Take long enough - about 60 scans an afternoon..that's for slides and negs. I took a couple in TIFF mode and got a huge file - wondered what I' d done. Canon 9000F scanner. I use Vuescan instead of the Canon scanning program which is crap.
- He felt that his whole life was some kind of dream and he sometimes wondered whose it was and whether they were enjoying it.
I'm still an amateur at it all but get good results at 800 dpi for full screen viewing. I printed one from a negative on photographic paper,at A4 and it came out surprisingly well. It very much depends on the quality of your original, whether it be a negative, slide or photo. I think most of my slides were taken with an SLR. Yesterday, I found a small black and white photo of me as a 17 year old. It was about 6cm on its longest side and looked in good nick. I scanned it at 1800 dpi and whilst it was ok at postcard size, it really showed surface defects at bigger magnifications.
As virtually all our looking at our old photos will be on a normal computer screen, I'm happy with 800 dpi, partially being pragmatic because of the time it takes, which you mention. Not really interested in even casting them to the TV.
The tinkering about with old photos, slides and negatives continues! We have a lot of black and white photos, some going back to the late 1800's. A lot of these photos are quite small and it would be nice to enlarge some to hang on the wall without getting pixellation losses. Ditto if I wanted to crop part of a photo and enlarge it. An acquaintance suggested that I look at a program called Topaz Gigapixel AI so I downloaded a 30 day free trial. Bloody hell, it's impressive.
The original photo of my grandmother was taken in 1949 (ummm... I was 2 ). It measures approximately 5.5 cm x 8.6 cm. I scanned it at 300 dpi then cropped it to just cover Gran's head and shoulders which had an original size of about 2cm square, then ran it through Gigapixel at a moderate enhancement. You can see the improvement on the side by side screen shot of Gigapixel (left original, right enhanced). At lower levels of magnification, it looks even better. The 1975 colour cropped photo of my wife with our Escort is about 3.5 cm square on the original photo. It was just a cheap camera so quality wasn't flash to begin with but at even moderate enhancement, the improvement is noticeable; particularly the trees in the background. We're limited by what can be reproduced on KB but you get the gist. There's a good video of its capabilities here: https://www.youtube.com/watch?v=2khOr_FqykA . There's also a version for enhancing old videos.
At US$80, it's a really useful tool for cropping and/or enlarging without the losses you'd normally get. I think I'll be buying it when the trial is up.
Looks like what they used on CSI.
I scanned some old black and white photos recently for my mother's funeral. They looked OK but I'm sure there's more detail in them, so I'll trial Gigapixel.
I also discovered that my smartphone camera did a better job of colour prints with its 64 MP resolution and AI than the scanner. The world is getting fairly crazy...
There's some cool stuff on the market now and the Gigapixel free trial is a great way to check it out. I've just scanned a B&W photo for our local goldmining museum of a group of people taken in the early 1900's. The original is smaller than postcard size and facial detail is poor. After running it through Gigapixel and printing it on A4 inkjet photo paper, facial features are incredibly clear. Good luck with trying it out!
That brought back memories. Yes, they could do "anything" on CSI .... 8-)
Like:
Tell you that the drop of engine oil - found at the crime scene - belonged to a BMW sedan manufactured between years X and Y
or
Tell you that they had extracted some DNA from a mosquito that had bitten the victim, and could recover the victim's DNA. [ Why they would ever want to do so - when you had the victims body separately available for taking of samples - is besides the point ]
These days, they are even more sophisticated.
They have "security services" that can hack into a remote computer system of any third party and immediately find exactly what data they are looking for .... 8-)
Sorry to intrude on your thread, Blackbird.
Oh God - did you ever watch CSI: Cyber? It was some of the most unintentionally funny shit ever.
There was one episode where they made this massive deal about the IP address (and specifically, an IPv4 address) because this could be linked to the Killers Address (Which isn't true in the way they were portraying it)
But even funnier was the fact that the IP address wasn't even a valid IPv4 address (which, due to to each octet being 8 bits long, can't have a number greater than 255) - so each time they showed up the IP: 951.27.9.840 - I had a great laugh.
Physics; Thou art a cruel, heartless Bitch-of-a-Mistress
As a final example, there's an alternative AI algorithm on Gigapixel AI for enhancing what they term man-made objects as opposed to the human form. It enhances straighter edges etc. Just tried it for the first time and got pretty remarkable results. I took the photo in 1968 and the dimension of the photo area shown below is about 3 cm square on the original. Left is the original enlarged and right is after Gigapixel has done its thing.
For interest, it's one cylinder of Alf Hagon's JAP-powered V twin drag bike - the first British bike to break the 9 second quarter mile barrier in 1967. It also achieved 207 mph on an RAF runway a couple of years later. Not bad for an unfaired bike at that time
Impressive software capabilities. I suspected that 'hard' items in photos as opposed to things like faces and fabrics worked better as I noticed the buttons in your grandmother's photo were particularly enhanced compared to her face and hair.
Better data for the software to work from I guess. If the smarts of the software are along the lines of what I suspect it will be able to recognise hard lines better than random curves and detail. In one of my previous jobs I processed 3D laser scanning data and the software's ability to 'build' pipework and other engineered items was incredible. Rendering a curved surface such as a yacht hull was far more challenging however as no straight lines or true radii for it to 'get traction' from.
That's pretty much right on the money. Since that post, I've played about a bit more and have been getting some good facial recognition results, Obviously the better the original, the better the final result but I recently enlarged a photo from school days where all the pupils were lined up in rows for the photographer. Each pupil's head would have fitted in a 5mm square on the print I have. After scanning it and running it through Gigapixel, each head would have been about 20mm in darned near perfect clarity.
That laser scanning you mention - I've seen renderings on a TV programme where the Paris catacombs were laser scanned and turned into 3D images. The level of detail is unbelievable. Must take a lot of computing power to process.
Processing laser scan files definitely puts the computer to the test. Dealing with many millions of x,y,z coordinated points to start with but then recognising their positional relationship with one another to digitally model pipework, flat planes, RSJ beams, etc is very impressive to see being processed.
For architectural or topographical subject matter a scan density of 20-50mm (horizonally and vertically) is sufficient but to gain data sufficiently dense to accurately model engineering projects can require 2-5mm 'resolution'. For the likes of a 75m long (tall) refinery tower I scanned which had 80+ connection flanges on it the file size was well into GB territory, which was then used to generate 3D AutoCAD modelling that confirmed all flanges were within +/- 2mm manufacturing tolerances including bolt hole orientation. It is really impressive technology!
This video gives an idea of what is involved in gaining and processing 3D scan data:
After making copies using a Canon 80D with bellows & lense using natural light, then converting to positives in software, I found with the sheer number of slides & negatives I wanted to scan I needed lockdown to be 2 years. I ended up purchasing a Wolverine F2D Saturn scanner, a stand alone unit that I connected to a 50" flat screen & saved them onto a SD card. Although the quality is not as high I was able to identify which pics were worth reproducing in higher resolution with the camera & bellows. The Wolverine also scanned a greater variety of negative sizes. Very easy to use & good for the majority of negatives. The most important trick is to number the negatives to match the image numbers.
Suck, Squeeze, Bang, Blow arent just the 4 cycles of an engine
Good stuff! It's a never-ending job, isn't it? At least it being winter has helped a bit. I ended up scanning quite a few poorer quality slides because the content still held good memories. Post-processing software helped in some instances to improve them. I still have slides from the 1969 Isle of Man TT and the 1970 Transatlantic Match Racing series to find. Really looking forward to those coming to lightt!
There are currently 1 users browsing this thread. (0 members and 1 guests)
Bookmarks