Jeeez, this was a really informative video. I don even own a DSLR scanning setup(I'm rocking an Epson flatbed), but this video was still a great watch.
amazing nerdy review! i just about pull the trigger on a r4, and I have been using nikkor 60 on a first gen a7. this is exacally what I need to see. it covers everything i didn't already know.
I thought pixel shift was a gimmick until I tried landscape photography with it on the Lumix G9, and I did one regular and one 80mp image, and honestly, its shocking how VAST the difference is, if you got a sharp lens like the Sigma are really good for this, then you be surprised how insane, there is literally 4 times more details, its much more detailed than my 5DSR is with its 50 megapixel sensor. Kinda mind boggling to have a medium format camera in a m4/3 sensor size. Its only working with very static things however.
Hey, nicely done! Echoes what I've seen with my G9 and the high res pixel shift mode. I have only DSLR scanned a few times for curiosity sake (then lost interest to explore further) as I normally use the Coolscans. However, have you: 1. Tried focus stacking? For those without high quality film holder and fixtures, this should make life a whole lot less frustrating. No worries about film curl or precise orthogonality between the optical axis and scan plane. It should also lets you use the macro lens without concern with depth of field. 2. Tried strobe as lighting source? Won't work with pixel shift obviously, but should make for a faster workflow as darkening the room for exposure wouldn't be necessary. 3. Tried DIY infrared dust removal. Normal light exposure then a reference exposure with an IR source (long enough or strong enough to let some through the sensor's IR filter). You sound like you have an engineering background, so a bit of ImageMagick scripting or OpenCV should about do it to interpolate through the IR reference frame.
My kit has changed a bit since this video. I have a Nikon coolscan lens from the 8000 and a 99 cri light source. Just need to get the pixel shift camera. The s5 is the top runner right now
I got my Oly EM5 Mark 3 for negative scanning, 20MP sensor and can scan 80MP. While I mostly shoot film, when I shoot digital, I like MFT for the compactness (since it's usually for holidays).
I've got a GH5 and it's great for when traveling light but no pixel shift that I'm aware of. My A7RIVa is a beast for photos but stupid heavy with good lenses so it just depends on what I'm taking pictures of on which one I use. For video the GH5 way more often than the Sony.
@@ShinyTechThings I had a gx9 that I was using for scanning, but traded it for the em5 for pixel shift and weather sealing. On the whole it was a great choice I prefer oly JPEGs out of camera (though I much prefer the Lumix grain simulation)
Awesome video! I just got a Sony A7r5 and was looking forward to trying to take advantage of Pixel Shift for larger formats but didn't think to use it for small ones. Actually, Ken Rockwell claims that the lower megapixel settings (e.g. the in camera native 26mp setting) eliminates Bayer Interpolation. Did you ever try out the Medium size images and compare them to the 24mp nikon ones? That would be a huge savings in the "don't have to pixel shift 35mm then downscale* workflow
What I found was using the pixel shift 60mp mode for full colour and downscaling the tiff using imagemagik to 32mp works really well and helps average out the noise in sensor and grain.
@@SprocketHoles I have also been wondering for a long time whether it is possible, by shining a color negative through red green and blue glasses, and combining three photographs into a single one, to get a color negative scan at the output. idea op type foveon matrix only multi megapixel
What a great presentation! I'm contemplating getting a A7R4 and wonder how long does the camera usually take to gather the 16 exposures? I've only been able to take one 16x pixel shift picture at the camera store. And that was nowhere near the 10fps the camera advertises. It took 8 seconds for the camera to take the 16 shots. My Lumix G9 does the pixel shift thing in about a second. I'm now wondering if I got a setting wring or if the Sony is really so slow?
@@SprocketHoles Fantastic drilldown. I even went further. I strapped a Sigma 105mm Art to my EOS 3 and shot a picture of a 1951 USAF resolution chart on Adox CMS 20 II. I digitized it using the Sony with 90mm macro. Group 7 Element 3 was resolved, which is 8193 dpi so ~90mp! I now use an Olympus em5ii w/ TTArtisan 40mm Macro as a dedicated "scanner". The lens and camera cost a total of $400. I get 64mp raw scans in 4:3 format of my 6x8 and 645 negatives, and the Olympus tethering works great. Cheers!
I am using the Sony Remote software with a 4 shot pixel shift on the A7R 5. Lightroom now handles the ARQ merged file that Remote creates, and I manually merged the ARW files into a DNG with PixelShift2DNG. Side by side, they appear to have done a very similar job in merging. I do not see any sharpening happening on the Sony Remote side (ARQ file) than I do on the DNG. I am wondering if the Sony Remote software doesnt' do the sharpening like Image Edge does, even though it's part of the same software package.
I am willing to share the ARQ merged file from Sony Remote and DNG from PixelShift2DNG for anyone who wants to compare. Let me know where I can share it. =)
I think you're not quite accurate with how the interpolation works in demosaicing. Of course there are many algorithms. What you describe is a basic algorithm that I don't think is used much. More specifically, the channels are not interpolated independently of each other. Rather, an assumption is made that for a group of pixels, the hue is the same, which works pretty well for human vision. With this assumption, the other channels are used as part of the interpolation to calculate the brightness.
I know it was a basic method and the method used is a lot more complex. But no demosaicing method can recreate an perfect random noise pattern so errors will be introduced. That what I was getting at.
Nikon scans look much more colorful and interesting. What is the point of increasing the scan by 200 percent or more, considering its sharpness, if the general idea of a digitized Sony image is worse than that of Nikon. at sony, they seem to be flat, without small tones and transitions.
Really loved how in depth you went with the diagrams and such, excellent video!
Impressive effort, very informative. I think CZcams compression ruined some of the details but your accurate descriptions resolved that issue. Cheers!
I enjoyed your video. It's nice to hear you being technical on those concepts.
This video is absolutely wonderful. thank you very much for this content.
Maximum effort in explaining this. Great job!
Now Fuji GFX 100s with pixel shift! Thank you for your review, very informative!
I would need a much better lens for that. Maybe a Scanner Nikkor
I've been thinking about this and my theory is that it could work. I have the A7RIVa so I'd love to see how these come out.
Jeeez, this was a really informative video. I don even own a DSLR scanning setup(I'm rocking an Epson flatbed), but this video was still a great watch.
amazing nerdy review! i just about pull the trigger on a r4, and I have been using nikkor 60 on a first gen a7. this is exacally what I need to see. it covers everything i didn't already know.
Great! Thank you for helping me gain more knowledge about pixel shift.
3 min in. On the edge of my seat!
Fantastic video mate! Thank you!! Only minor addition I would have loved is to include the lens used to take the actual shot. Thank you!
I thought pixel shift was a gimmick until I tried landscape photography with it on the Lumix G9, and I did one regular and one 80mp image, and honestly, its shocking how VAST the difference is, if you got a sharp lens like the Sigma are really good for this, then you be surprised how insane, there is literally 4 times more details, its much more detailed than my 5DSR is with its 50 megapixel sensor. Kinda mind boggling to have a medium format camera in a m4/3 sensor size. Its only working with very static things however.
Great input. Could you please add the hardware description that you used to hold camera and to hold the negatives?
Works great on the Panasonic G9 ii
Hey, nicely done! Echoes what I've seen with my G9 and the high res pixel shift mode. I have only DSLR scanned a few times for curiosity sake (then lost interest to explore further) as I normally use the Coolscans. However, have you:
1. Tried focus stacking? For those without high quality film holder and fixtures, this should make life a whole lot less frustrating. No worries about film curl or precise orthogonality between the optical axis and scan plane. It should also lets you use the macro lens without concern with depth of field.
2. Tried strobe as lighting source? Won't work with pixel shift obviously, but should make for a faster workflow as darkening the room for exposure wouldn't be necessary.
3. Tried DIY infrared dust removal. Normal light exposure then a reference exposure with an IR source (long enough or strong enough to let some through the sensor's IR filter). You sound like you have an engineering background, so a bit of ImageMagick scripting or OpenCV should about do it to interpolate through the IR reference frame.
My kit has changed a bit since this video. I have a Nikon coolscan lens from the 8000 and a 99 cri light source. Just need to get the pixel shift camera. The s5 is the top runner right now
Thanks a lot for this very instructive and useful video.
Excellent video. Tons of knowledge. THANKS!
The Sigma cameras with Foveon sensors might actually make for good scanner cameras. Although they're APSC, they don't have Bayer filters.
Fantastic breakdown.
I got my Oly EM5 Mark 3 for negative scanning, 20MP sensor and can scan 80MP. While I mostly shoot film, when I shoot digital, I like MFT for the compactness (since it's usually for holidays).
I've got a GH5 and it's great for when traveling light but no pixel shift that I'm aware of. My A7RIVa is a beast for photos but stupid heavy with good lenses so it just depends on what I'm taking pictures of on which one I use. For video the GH5 way more often than the Sony.
@@ShinyTechThings I had a gx9 that I was using for scanning, but traded it for the em5 for pixel shift and weather sealing. On the whole it was a great choice
I prefer oly JPEGs out of camera (though I much prefer the Lumix grain simulation)
thank you
Hi, Thanks for this video.
Do you still prefer Epson V850 scanner or you now have better results with camera scanning?
Awesome video! I just got a Sony A7r5 and was looking forward to trying to take advantage of Pixel Shift for larger formats but didn't think to use it for small ones. Actually, Ken Rockwell claims that the lower megapixel settings (e.g. the in camera native 26mp setting) eliminates Bayer Interpolation. Did you ever try out the Medium size images and compare them to the 24mp nikon ones? That would be a huge savings in the "don't have to pixel shift 35mm then downscale* workflow
What I found was using the pixel shift 60mp mode for full colour and downscaling the tiff using imagemagik to 32mp works really well and helps average out the noise in sensor and grain.
I would love to see BW scans done with a Leica M10 Monochrom + Macro-Elmar. No bayer-filter to worry about. Quick and easy full 40mp in theory?
Yep in theroy
@@SprocketHoles
I have also been wondering for a long time whether it is possible, by shining a color negative through red green and blue glasses, and combining three photographs into a single one, to get a color negative scan at the output. idea op type foveon matrix only multi megapixel
@@dzhuk Maybe for slide film but not for colour negative film as the dyes used are the inverse (cyan, yellow and magenta).
@@SprocketHoles yes, that's right, I was wrong, but I tried to convey the principle itself)
What a great presentation! I'm contemplating getting a A7R4 and wonder how long does the camera usually take to gather the 16 exposures?
I've only been able to take one 16x pixel shift picture at the camera store. And that was nowhere near the 10fps the camera advertises. It took 8 seconds for the camera to take the 16 shots. My Lumix G9 does the pixel shift thing in about a second.
I'm now wondering if I got a setting wring or if the Sony is really so slow?
It took about 3 seconds for me.
@@SprocketHoles Thank you!
@@SprocketHoles Fantastic drilldown. I even went further. I strapped a Sigma 105mm Art to my EOS 3 and shot a picture of a 1951 USAF resolution chart on Adox CMS 20 II. I digitized it using the Sony with 90mm macro. Group 7 Element 3 was resolved, which is 8193 dpi so ~90mp!
I now use an Olympus em5ii w/ TTArtisan 40mm Macro as a dedicated "scanner". The lens and camera cost a total of $400. I get 64mp raw scans in 4:3 format of my 6x8 and 645 negatives, and the Olympus tethering works great.
Cheers!
Excellent video. Thanks! Can you post the bash script you briefly showed for the down sampling with imagemagick?
Does this affect your relationship with the V850? If so, would it be your prefered solution for larger (4x5) negatives now?
I sold the v850 months ago. I'm back on the DSLR scanning train.
what lens are you using?
I am using the Sony Remote software with a 4 shot pixel shift on the A7R 5. Lightroom now handles the ARQ merged file that Remote creates, and I manually merged the ARW files into a DNG with PixelShift2DNG. Side by side, they appear to have done a very similar job in merging. I do not see any sharpening happening on the Sony Remote side (ARQ file) than I do on the DNG. I am wondering if the Sony Remote software doesnt' do the sharpening like Image Edge does, even though it's part of the same software package.
I am willing to share the ARQ merged file from Sony Remote and DNG from PixelShift2DNG for anyone who wants to compare. Let me know where I can share it. =)
My guess would be that the image edge software was cranking the sharpening on me. maybe its chsnged since I made this video.
Do you think this system could have the same quality as a Drum scanning?
With the right lens, I think it will beat every scanner out there. I do have that lens but Im waiting on some parts atm.
I think you're not quite accurate with how the interpolation works in demosaicing. Of course there are many algorithms. What you describe is a basic algorithm that I don't think is used much. More specifically, the channels are not interpolated independently of each other. Rather, an assumption is made that for a group of pixels, the hue is the same, which works pretty well for human vision. With this assumption, the other channels are used as part of the interpolation to calculate the brightness.
I know it was a basic method and the method used is a lot more complex. But no demosaicing method can recreate an perfect random noise pattern so errors will be introduced. That what I was getting at.
I’m a former film shooter with a a7r4. Unfortunately technology has gone beyond human vision.
Nikon scans look much more colorful and interesting. What is the point of increasing the scan by 200 percent or more, considering its sharpness, if the general idea of a digitized Sony image is worse than that of Nikon. at sony, they seem to be flat, without small tones and transitions.
Sorry but I can't see a difference. I'll stick with my Nikon 24MP & 45MP scanning.
Same here. The difference is very negligible. And I own a7rv.