Making Big Prints from low megapixel images - resizing and sharpening (A2 from 6.6MP)
HTML-код
- Опубликовано: 29 авг 2024
- Keith looks at examples of processing raw image from an old camera with modern software.
Covering aspects of sharpening, RAW processing and up-sizing for a print image, with examples showing just how much current software can surpass even the best available at the time the photo was taken. The part of the image enlarged to A2 is actually a 6.6MP crop of the original 11MP Canon 1Ds image
Keith has written many detailed articles and reviews looking at all the different aspects covered in the video.
This video accompanies a more detailed article at:
www.northlight...
The prints of the mountains and aspens are discussed at:
www.northlight...
There are free demo versions of Sharpen AI and Gigapixel AI available via the articles below.
The original Topaz Gigapixel AI review with lots of examples discussing RAW processing is at:
www.northlight...
The current version of the software is looked at in:
www.northlight...
Sharpen AI is covered at:
www.northlight...
and
www.northlight...
-----------------
If you'd like to make a small donation, I have a Kofi page:
"Buy me a coffee" ko-fi.com/keit...
-----------------
My articles and videos are always free to access.
Any help with running this channel is gratefully received.
-----------------
I also have some affiliate links which earn me a commission if used.
US Amazon photo/print gear: amzn.to/3l9vJC6
B&H Photo: www.bhphotovid...
Adorama: www.adorama.co...
Great video, thanks. What happens to the original RAW file in this workflow? Do you first use Topaz and save the picture as a specific format e.g. TIFF? Which is then used in PS or any other editing software? Cheers
Thanks - The raw file is processed normally, but do note that this video is only a supplement to the original more detailed written article:
www.northlight-images.co.uk/upsizing-and-sharpening-for-making-a-print/
The articles are where the real detail goes [not written for reading on mobile devices BTW]
The prints of the mountains and aspens are discussed at:
www.northlight-images.co.uk/printing-low-megapixel-images/
Thanks for the well demonstrated and explained process. It's interesting that you mentioned the sensor dust issue that has plagued most digital cameras since day one. I have been shooting Olympus (now OM System) cameras for 20 yrs. -- professionally and personally. In all that time, I have never once had to clean my sensor. Olympus introduced the first sensor dust cleaning function (the SSWF) in the E-1 Model and has steadily improved it ever since.
Thanks - it's definitely got better over time, but bigger sensors do have more of a chance for dust to appear.
Modern upscale and AI tools are basically magic. I've printed 5MP images at 18"x24" and they look good good, I don't pixel peep the prints but from the proper viewing distance, 4-5 feet, they look great. Also printed on canvas, and it takes away any need for intense intricate sharpness, the texture of the canvas makes it softer anyways. I'm loving this new age of editing, has made photos much more fun to play with. I'm also a die hard photoshop user, I can't imagine using anything else that has more limitations.
Yes, the results can be astounding, if you've worked on smaller images in the past
I have older photoshop also and this video was so down to earth compared to other cookie cutter You Tubers. Very easy to understand. Thank You!
Thanks - Glad it was helpful!
I sometimes rent a holiday cottage that has photographs of local scenes on the wall. They are large, about 3 - 4ft in size, were probably taken on an old phone or very early digital compact camera, and printed on canvas. The sensor must have been in thousands of pixels, not millions. They are pretty bad technically, but I'm probably the only person to notice. A 6mp camera would produce an image light years from these shots, never mind a modern 24 mp version - which is already viewed as a bit low for "serious" photography according to some. It's marketing, pure and simple.
Yes - When I next get a big printer here I'll be doing some experiments about this - last time I had one here was before I started adding videos to my written reviews/articles.
I am always keen to try out your concepts, but, alas, I don't have enough room for all the big prints. Thanks for your great videos and reviews.
Thanks - the testing work produces a lot, but a few do make it to the walls ;-)
Fascinating stuff again Keith.
Strangely appeared not long after I had been watching an interview with Tony Levin who was saying his biggest regret, with hindsight, was switching from film to digital to quickly, having a period which produced thousands of images he considers "Unusable."
All good stuff again.
Thanks - the 1Ds was my first serious DSLR in 2004. My best move was, right from the start, to shoot in RAW and JPEG and to keep both in the archives.
Fascinating process Keith, I'll have to save this and watch it again while I practice. Thank you.
Thanks - there is more detail in the linked articles
Very helpful video Keith, packed with useful information, particularly the bit about the extra detail coming through in prints acting as a form of sharpening. I assume this negated the need for any further output sharpening in that particular case. Thanks for sharing this info.
Thanks
The necessity for print sharpening does seem reduced - or perhaps I should say modified. I've still not experimented enough to say for sure. This is something I want to test more when I next get a big printer here.
Thanks, sir. Funny is that we never worked together but you have mostly confirmed my workflow and intuition. The rest is actually that expertise (or "way of expert thinki g") you share.
TIP: for both problems with sky you have mentioned, few quick settings in Darkrable and you get everytning done. First - you can choose demosaic alghoritm. Every picture can work best when demosaiced with particular method. And, then you use "parametric mask" when you set mostly blue colored areas with saturation between x and y, within part of tonal range between x and y, and some more params if it's needed. And you sharpen all but this masked area, plus add denoise only in chroma channel, not touching luminance noise (which is responsible for details). As simple as analyze and less of minute of setting sliders.
Thanks - I will have to have another look at DT ;-)
I just have to get past the fact that every time I've tried it, I've rapidly felt like I've travelled back in time to my UNIX sysadmin days... [unfair, but it's why I've never reviewed many bits of software that felt alarmingly close to developers...]
If you get an advantage from it, let it be some reward for sharing your printing expertise. Complete university course with written materials and lectures in form of videos.
I've moved to using Gigapixel AI for my old 12mpx images. I find the results superior to other options, and the ability to tune the scaling based on the type of shot is invaluable. I have a 36x24 inch print using one of the 12mpx source files and you'd never know it.
Yes, especially if you tweak the source image it is using. Quite small changes can make noticeable differences in the final results
@@KeithCooper Oh, GP-AI also plays extremely well with my film shots, which many other options had issues with, even other products by the same company.
That's good to know - a lot of software has difficulty with grain.
@@KeithCooperI've gone back to film more the last few years, so that was a requirement for me.
With the 108mpx on my S21 Ultra plus having raw format, I haven't been using my EOS R much except for birding and low light. The results between the two are impossible to see a difference, even at A2. With portrait mode AI getting so good, it's truly difficult to see the difference for people shots. They're just pouring so much R&D into it, I can see a future not far off where I don't even own digital FF kit anymore. I know I won't miss the weight😁
Good points - My architectural work and the tilt/shift focal lengths available pretty much mandate a bigger sensor. That said I got the 1Ds out of the drawer when I was doing this video and it was a hefty old beast - I still look forward to the 5Ds mirrorless replacement ;-)
Excellent lesson. Learnt a lot from your thought processes and work flow. More of the same please.
Thanks - I'll see what I can come up with...
I assume the image shown is cropped as 3144 x 2096 is 6.6 MP and the 1Ds sensor was 4,064 × 2,704. So the example in the video is starting with a 6.6 megapixel image which is more impressive than starting with an 11 megapixel image.
Yes - thanks for spotting that.
I worked this out after I'd finished the video, and forgot to update the text!
It does make it more impressive ;-)
Thanks for this. Great information. I also enjoyed your description of a camera's build quality in terms of it's ability to be weaponised. Oh and your tea's gone cold.
Thanks - Fortunately I'm not a hot tea person ;-)
I shoot still images from a 12mp drone and I'm currently playing with Gigapixel as a way of boosting that. So far I'm very impressed with it.
Yes, it's very good - for low MP images, it really shows the benefits of good lenses.
The key for my older smaller images seems to be getting the source image processed in a way which aids the enlargement process
❤ Thank you Keith, your videos ar3 an inspiration.
Thanks - with the P5300 here for testing, it's time for a few more large prints ;-)
That looks like Seaside, Oregon. I used to live there. I love your videos, so helpful!
Thanks - I've not been back for a while, but will definitely do so!
I've quite a lot more pics from the region, from these articles, long before I started doing the videos
www.northlight-images.co.uk/keiths-photo-travels/
This is a really useful video Kieth. Thank you👍
Thanks - Glad it was helpful!
Topaz products are fantastic.
Yes, some impressive results ;-)
7:18 the other day I did an interesting little test, I printed a [1800 x 1200, 6" x 4" 300 dpi] image, shrunk it to [900 x 600, 6" x 4" 150 dpi], printed that, then upscaled it back to [1800 x 1200, 6" x 4" 300 dpi] using gigapixel ai. The 150 dpi image was clearly lesser quality but some parts of the upscaled image was better than the native 300 dpi
I love how new editing techniques and new software keeps me going back to my old raws :)
Yes - excellent test.
It's why I keep those raw files...
hehe always keep the raws. I have one folder with all my raws in chronological order in sub folders based on date. And a separate folder with all my edited images, including copies of the raws if I change them in any way. Easy to find what im looking for and I dont have to worry about messing up my raw files.
Very helpfull and interesting Video! I like your Tutorials very much. I learn very much from it. Thank you and i wish you and Karen Happy Easter
Thanks - All the best to you too!
Even though I’m into astrophotography I feel fortunate to find this video. I use dedicated Astro camera. I use pixinsight not photoshop to process with. I use bay photo for metal prints. The scale at 25 dpi . I would buy topaz gigipixel for upscaling. But maybe photoshop as it can improve the image . Thanks for the video.
Glad it was of interest. The characteristics of many astro images means that often sharpening and resizing needs subtle changes in the approach to sharpening - principles are similar, you just need to experiment a bit more with settings ;-)
Hey Keith - newcomer here. I really like your very informative content. Thank you for the tips!
I really want to see your showcased pictures in high resolution.
Now, I don't know about your editing process for these great videoes, but if you "upscale" by exporting in a resolution 3840x2160p and bitrate of minimum of 35 mbps - you can gain way crispier video. Even though it is shot in 1920x1080.
It has something to do with the youtube formatting when processing videos.
I hope you can use my tip - Let me know if you need any assistance ;) - Casper
Thanks - High res images are generally kept for the written articles ;-)
As to videos...
I've been doing the YT videos for less than 2 yrs (initially at the behest of Canon), and the articles for nearly 20.
All the video stuff I do at the moment (using DV Resolve 17) takes long enough and eats up my disks :-) - maybe when I get my office cleared and am not doing the videos in the corner of the kitchen, I'll look again at the process (thanks) ;-)
Not working on a 2012 Mac Pro might help, but I'm afraid that unless I win a lottery, no new kit is turning up any time soon.
Bravo pour vos videos, continuez dans cette voix, i am more an artist than a technic,.also I come to buy a printer: le sc p700 epson, and your videos are a good help for me. cordialement
Glad to help
Thanks for the video Keith, when you use gigapixel do you sharpen ( with Photo AI) before or after ? Does it matter? Thanks Rick
Good question - I've found it depends on the image and the size [it's with Sharpen AI - I've not actually tested Photo-AI yet].
See the original article for this video
www.northlight-images.co.uk/upsizing-and-sharpening-for-making-a-print/
See the links in the article to other articles - this article pre-dates my making of videos.
PS My videos are often secondary to more technical articles - I've been writing the articles for ~20 years - the videos for not yet 3 years ;-)
great video, appreciate the time.
Thanks - glad it was of interest!
Fantastic vid as always , very interested to know about Gigapixel AI as i have a load of images i love taken with my Canon 5d (classic) in which id love to print A3+ size on my A3+ size printer, thank you for sharing this knowledge.
I print 12 MP images upto A2 (four A4 sized, or 2 A3 sized, images taped together) and people have a hard time finding the glue lines... I do not use ANY software for 'improving' the images. AI or not AI, no software will be able to predict what's in a pixel that is not there. Just give it a try.
I cannot post links to my website.
"no software will be able to predict what's in a pixel that is not there" - yes seems intuitive, but increasingly wrong I'm afraid. We are at just the start of this process. Personally I'll take any tool available to 'improve' images if it makes for what I feel is a better print.
Yes, with care, I can print A2 from my old 11MP images - I have some on the wall from a 2005 exhibition. However, if I was to remake one of these prints. I would use modern software and techniques.
Thanks - the 5D sensor is pretty much the same as in my 1Ds. This works really well for images at A3+
Thank you
Thanks - glad it was of interest
Hi Keith, I'm a great fan of your Videos. My Englisch is not so good.Have you used only Gigapixel AI for this photo or also Sharpen AI?
Use Google translate on this!
www.northlight-images.co.uk/upsizing-and-sharpening-for-making-a-print/
Keith, another great video. I’m a Capture One, not a Photoshop user.
What in your described conceptual process will I be unable to do given that C1, unlike LR, allows using layers/masking, etc.
Thanks👍🏻👍🏻
I've not used C1 for several years, so can't say for sure.
The essence of the process covers
RAW conversion/sharpening/resizing/adjustment/printing
Now, you may need to export images and run them via the standalone topaz apps, and then import back to print.
Seems potentially clunky, but in principle it should work ;-)
helpful, thank you.
Thanks - I'm looking forward o getting a big printer to test for a change ;-)
Thank you for another great video, ohhh ... i think your coffee is getting cold ;-)
Yes (tea) - and it had 16 seconds in the microwave to revive it ;-)
Sharpening can add in some areas and drain the life from an image in other areas, as it did with the bird's peak and feathers which improved at the expense of draining life out of the eyes.
No - Actually handling the physical print - I disagree. Unfortunately, a compressed RUclips video is a very blunt tool for critical sharpening and print detail observations.
More to the point though, this is an exercise in looking at aspects of a process and how modern software gives sharpening and re-sizing options that simply did not exist a few years ago.
This video is primarily about the process and options it gives in editing an old image... Sharpening is not a thing on its own, it's an element [or elements] in a process to be used as needed/desired.
The details are discussed in the main [written] article at:
www.northlight-images.co.uk/upsizing-and-sharpening-for-making-a-print/
and in
www.northlight-images.co.uk/printing-low-megapixel-images/
But hey, it's a seagull taken on a digital camera nearly 20 years ago, so I'm pretty happy with it as an A2 print... :-)
@@KeithCooper I appreciate your replying to my comment.
Did you crop the bird in the original photo?
Yes, that makes it an even bigger enlargement
Could I just comment that the two pieces of Ai software you use are very expensive ,are there alternatives?
Undoubtedly there are other ways - this is the approach I'm likely to use for my commercial work.
See this article from a few years ago where I took an 11MP image up to 3m x 2m
www.northlight-images.co.uk/making-a-giant-print-from-an-11mp-image/
It goes through a lot of the process in considerable detail - Should be adaptable to a tool like Affinity Photo, and excellent and relatively cheap image editing tool.
Both my D90 and D7100 produce 12 MP images which I can easily print at A2 size. A1 is also still good as long as you keep a few feet distanc from the prints. And I use Out Of Camera JPEGs.
Yes - it's possible, but I prefer the option of creating higher res version, if it makes for a good print.
@@KeithCooper At what point does it become 'painting' instead of photography?
Ah that old chestnut... ;-)
It is a false dichotomy IMHO
If you want to be concerned about 'invented' detail, then what about the colour pixels interpolated and 'created' when demosaicing a RAW camera file.
It's a personal choice and depends on what the print is for.
I find an HDR style image far more unpleasant than a big print with 'made up' fine detail.
Would I use a tool like APG if I worked in photojournalism, or photographed crime scenes - no. However I do neither and my aim is generally a print which looks right.
Will I label images as in somehow 'fabricated' nope, not a chance ;-)
@@KeithCooper The demosaicing is indeed the first step where guesstimation is used.
Just updated to a D700 12 Mpix fron my broken down D300.I'm saving for a used D850. Ilike to be able to print 1m. I will not switch to Zed but I will get a Fuji GFX50sii.
Yes, that will help...
The Ai software you use are expensive,are there alternatives?
See this article from a few years ago where I took an 11MP image up to 3m x 2m
www.northlight-images.co.uk/making-a-giant-print-from-an-11mp-image/
It goes through a lot of the process in considerable detail - Should be adaptable to a tool like Affinity Photo, an excellent and relatively cheap image editing tool.
Do you ever use Affinity Photo? If so, what do you think about it?
I have experimented with it - I rather like it.
However, I've used Photoshop for years (remember - photography is my day job), so there is only so much experimenting I do ;-)
I may have a look at using it for a print next time I have a big printer here to test, but it won't be a detailed review.
@@KeithCooper Thanks. I found PS daunting and increasingly expensive. Still don't know my way around Affinity, but feel it is more friendly.
Have had an Epson Stylus Photo 1400 for years and made some (by my standards) good prints .... then it died. Struggled with fixing it and capitulated finally. Decided to spring for a new printer (no choice) and thought (economically) the ET-8550 would be fine ... and cheaper to print on. BUT they are not to be found! SO reconsidered the SureColor 700/900. More expense, to print on given the cost of cartridges, but perhaps better prints. Your videos are very helpful andI appreciate your clear and very straightforward analysis of the pros and cons. So, thank you!
Thanks - if you need any of the profiles mentioned in my [written] reviews - they are free for non-commercial use.
Interesting process, but in terms of purity, the "machine learning" is basically the computer painting new fake detail into the image.
It's art, you do what you want, but that's worth pointing out, imo.
Yes - If some notion of 'purity' means anything at all given the actual output from a Bayer sensor and how it's processed to get an image.
Modern Demosaicing algorithms incorporate machine learning based approaches.
Would you want to see the resurrection of Foundview and Trustimage?
Discussed more in my videos asking how much you are happy altering in an image
ruclips.net/video/xO0Lrd51hus/видео.html
Light room v Camera raw =
Not sure what this means?
I have a longstanding dislike of LR - but that's just me :-)
@@KeithCooper Good afternoon Keith. I am slavishly following your advice and recommendtions 😄 I've got as far as buying a Canon Pro-200 and boning up on 'how to' via your output. I was brought up short by your comment that using LR is 'like sending your films to a lab to be printed' rather than using PS 'having your own darkroom'.
Having started my career in photography in 1969 as Mr Copy Camera and progressing thru' the trad steps - film processing/darkroom/studio assistant etc I rediscovered photography courtesy of digital+LR after some years doing other things [don't ask!] . Using LR was like going back to wet darkroom work but with infinitely more possibilities.
My understanding of Adobe's creation of LR is that photographers, not needing the full-fat capabilities of PS, which is principally a graphic arts 'app', gave us the bits of PS that photographers need.
Being now ancient of days it is most unlikely that I will try to get my head round PS - I did a short 'interactive' course but most of it was irrelevant to photography. However, I took to LR like the proverbial duck ...
I know you would not have made that comment lightly so I would be grateful if you could expalin how you come to that view.
@@christophernation4793 I just dislike many features of LR - my caveat is always that this is very much a personal choice and not a recommendation!
Now, you did ask...
It is not a PS 'lite' - that's Elements ;-)
The forced use of a Catalogue was for many years an issue - I never wanted it. I have a dislike of the 'non-destructive edit' approach. There are no layers/masking/blending modes.
The printing is aimed to be easy to use at the price of control and subtlety - fine for snaps, not for large stuff... Of course that's a heresy for many - look at what all those US experts say... ;-)
If I was a wedding or event photographer, it might actually much better fit my required workflow, but I'm not one ;-)
LR has grated with me every time I've tried it - do what they think you want to do, it's fine, try and carve your own path, it hinders.
Of course, Adobe have also encouraged a whole industry of experts telling you how great it is, offering training, and nailing their colour to its mast.
I just say 'not for me thanks'...
Hi Keith, I'm a great fan of your Videos. My Englisch is not so good.Have you used only Gigapixel AI for this photo or also Sharpen AI?
A bit of both - it's covered in the main article
www.northlight-images.co.uk/upsizing-and-sharpening-for-making-a-print/
Maybe translation works better on my written work?