Ugly Hedgehog - Photography Forum
Home Active Topics Newest Pictures Search Login Register
Panorama
Why sweep and stitch?
Page <prev 2 of 2
Jul 6, 2018 12:14:39   #
Rich1939 Loc: Pike County Penna.
 
Tim Hoover wrote:
Glad to be of some help. If I may add to what I wrote before... When considering resolution there are really three things to consider, 1) The input resolution 2) the display resolution 3) the optics resolution.

The overall resolution is limited by the lowest resolution in this chain. So, no matter how many pixels your sensor has, if the lens can't provide the corresponding resolution, you have again wasted your money.

That said, I now shoot with a D850 (moving up from a D7100) and the improved resolution is dramatic. However, that is due to a combination of improved sensor and improved lenses. Does this show up in a reasonable sized print? Absolutely not (which you should know if you've been paying attention). However, I now have the option to make larger prints without resampling as well as cropping significantly if I need to, so there are some advantages.

And by the way, there are some very good algorithms available for upsampling (I'm thinking specifically of 'Perfect Resize') so if you want to print larger than the native resolution, you can do so with very little concern for lack of sharpness.
Glad to be of some help. If I may add to what I wr... (show quote)


What is a "reasonable sized print" ?

Reply
Jul 6, 2018 12:44:29   #
Tim Hoover
 
A 24MP image with no up or downsampling printed at 240 dpi is roughly 15x23".
A 45MP image with no up or downsampling printed at 240 dpi is roughly 23x34".

These are reasonably large for most people.

Good upsampling algorithms support somewhere around a factor of 3 without (very) noticeable loss of quality. The larger the print, the larger the viewing distance (in general) so the differences become moot.

Reply
Jul 6, 2018 15:56:28   #
Rich1939 Loc: Pike County Penna.
 
Tim Hoover wrote:
A 24MP image with no up or downsampling printed at 240 dpi is roughly 15x23".
A 45MP image with no up or downsampling printed at 240 dpi is roughly 23x34".

These are reasonably large for most people.

Good upsampling algorithms support somewhere around a factor of 3 without (very) noticeable loss of quality. The larger the print, the larger the viewing distance (in general) so the differences become moot.

I've been struggling with this and quite frankly I don't understand how you came up with those dimensions. As an exercise I wanted to change the DPI to 360, couldn't work it out. Maybe I need more coffee.

Reply
 
 
Jul 6, 2018 16:19:10   #
Tim Hoover
 
A standard sensor has an aspect ratio of 3:2
For a 45 MP sensor that is roughly 8216x5477 pixels (5612x3742 for 21 MP).

Divide those numbers by the printer's DPI and you get 34.2x22.8" for 45 MP and 23.4x15.6" for 21MP at 240DPI, (22.8x15.2" and 15,6x10.4" respectively) at 360DPI.

Reply
Jul 6, 2018 17:33:20   #
Rich1939 Loc: Pike County Penna.
 
Tim Hoover wrote:
A standard sensor has an aspect ratio of 3:2
For a 45 MP sensor that is roughly 8216x5477 pixels (5612x3742 for 21 MP).

Divide those numbers by the printer's DPI and you get 34.2x22.8" for 45 MP and 23.4x15.6" for 21MP at 240DPI, (22.8x15.2" and 15,6x10.4" respectively) at 360DPI.



Had another cup of coffee and the lights came on.
So, for a 24mp image you believe something approaching 11.1 X 16.7 would be about the right size for a reasonable sized print. Are you also implying that a larger size would degrade the image?

Reply
Jul 6, 2018 17:54:37   #
Tim Hoover
 
At that size one pixel on the sensor maps to one dot on the print (under ideal conditions where the printer puts out square dots of perfectly aligned ink). That is all the information available in the file.
If you print bigger, then an algorithm is employed to extrapolate the data to the larger size (effectively inventing data). As I mentioned previously, there are some very good algorithms available for doing this. So, would you NOTICE a degradation? I doubt it if you don't push the extrapolation too far.

Remember though, this thread started out in basically the opposite direction. That is, how many pixels are really needed for a given sized print? Many people seem to think that more is better. It's not. Any time you print SMALLER than the native resolution an algorithm is employed to throw away data so you gain nothing, and in fact risk degradation due to the algorithm employed to downsample.

Now, just on the unlikely chance that someone else has followed the thread to this point, let me emphasize that these resolution questions are all theoretical. Will you actually NOTICE a difference if you print a bit larger or small than optimal? Very likely not. Oh, and one last (technical) point, the term 'resolution' does NOT refer to the number of pixels in a sensor. However, everyone I've ever read on this forum treats them the same (including me) so I will continue with that sloppy practice unless someone call me out on it.

Reply
Jul 6, 2018 18:44:17   #
Rich1939 Loc: Pike County Penna.
 
Tim Hoover wrote:
At that size one pixel on the sensor maps to one dot on the print (under ideal conditions where the printer puts out square dots of perfectly aligned ink). That is all the information available in the file.
If you print bigger, then an algorithm is employed to extrapolate the data to the larger size (effectively inventing data). As I mentioned previously, there are some very good algorithms available for doing this. So, would you NOTICE a degradation? I doubt it if you don't push the extrapolation too far.

Remember though, this thread started out in basically the opposite direction. That is, how many pixels are really needed for a given sized print? Many people seem to think that more is better. It's not. Any time you print SMALLER than the native resolution an algorithm is employed to throw away data so you gain nothing, and in fact risk degradation due to the algorithm employed to downsample.

Now, just on the unlikely chance that someone else has followed the thread to this point, let me emphasize that these resolution questions are all theoretical. Will you actually NOTICE a difference if you print a bit larger or small than optimal? Very likely not. Oh, and one last (technical) point, the term 'resolution' does NOT refer to the number of pixels in a sensor. However, everyone I've ever read on this forum treats them the same (including me) so I will continue with that sloppy practice unless someone call me out on it.
At that size one pixel on the sensor maps to one d... (show quote)


I don’t know that resolution has ever been quantified. At least not in a way Joe Six Pack can make use of it.
In another vain I do know for my purposes I would rather have the pixels to throw away than need to fake them when wanting to go larger. With stitching and using normal or medium telephoto lenses my 24MP camera can give me a .dng file equivalent to a 36MP camera or larger. I also like the additional control that stitching allows over aspect and/or capture angles while diminishing distortion. My lens of choice is either a 50mm or an 85mm. Both used in the vertical position, not a whole lot of native distortion to deal with with those. While I have been blessed with a good eye for level when looking through the view finder, plumb (fore and aft) is problematic. Keystone correction fortunately is not a big deal in Photoshop.
FWIW I have printed pictures almost 2X what your formula would indicate is optimal and have been pleased with the prints

Reply
 
 
Jul 22, 2018 22:53:29   #
Gene51 Loc: Yonkers, NY, now in LSD (LowerSlowerDelaware)
 
jastewart wrote:
Conventional wisdom seems to be that to do a digital panoramic shot you sweep and stitch. Why? To make it harder to do? So you can't accurately frame what you are shooting? So you can blow it up to mural size? The latter is the only one that makes sense to me.

I recently came to digital after using a Hasselblad XPan, a wonderful rangefinder camera that allows you to take moderate panoramas in a single shot, with accurate framing and good resolution. I bought a Nikon D7500 after convincing myself that I could get similar results with it. I put a homemade mask over the viewing screen so I can see the framing for panoramic when I shoot. I then do a custom 1 x 2.74 crop in Lightroom. I maintain that the resolution I get is just fine for prints up to 20+ inches. Most importantly, you can previsualize the shot, which I have long believed is key for "art" photography (see Edward Weston, Minor White, Paul Strand, Ansel Adams, etc.).

I look forward to your comments.

John
Conventional wisdom seems to be that to do a digit... (show quote)


A stitched pano is a lot like a fine art painting of a landscape. The point of view and longer focal lengths possible with stitched images resembles what the painter used - his/her eyes and the muscles in the neck to rotate the head (and eyes) for a wider view. If you go to a museum, NONE of the classic landscapes "look" like they were viewed through an ultrawide lens. Ultrawides have a very nasty extension distortion and volume deformation at the edges. An interesting effect but it's something that has become passé since so many photographers have access to lower cost ultrawide lenses.

For the record - you can make a billboard sized image from an iPhone shot.

But the stitched pano does have the advantages of higher resolution, which can be used to reduce noise by downsampling.

This will tell you what image resolution you will need for specific viewing distances.

http://www.photokaboom.com/photography/learn/printing/resolution/1_which_resolution_print_size_viewing_distance.htm

Reply
Jul 22, 2018 22:56:02   #
Gene51 Loc: Yonkers, NY, now in LSD (LowerSlowerDelaware)
 
jastewart wrote:
Thank you for this. As I said, I'm new to digital and didn't completely understand the argument about native resolution and down/up sampling, but I get from it is that what I am doing isn't crazy and that the resolution I'm getting for the print sizes I'm doing would not be improved by more megapixels. I believed that there was a simple correlation between number of megapixels and resolution, but apparently it is far from that simple.

When I first thought about going digital, I rented a Nikon D850 (42 megapixels) and a D7500 (20.9). I shot the same scene at the same size with the same lens. I found NO difference in resolution of the images at high magnification, and of course I bought the D7500. So I tend to agree with you that many people have more megapixels than they need (and less money than they would prefer).
Thank you for this. As I said, I'm new to digital ... (show quote)


A print will show the difference. And if you crop you will still have more than enough pixels with the D850, more so than the D7500. And if you shoot in low light and high ISO, I think you will like the results from the D850 over the D7500. I like the D7500, but it is definitely not a D850 for so many reasons.

Reply
Page <prev 2 of 2
If you want to reply, then register here. Registration is free and your account is created instantly, so you can post right away.
Panorama
UglyHedgehog.com - Forum
Copyright 2011-2024 Ugly Hedgehog, Inc.