I wrote a response earlier to this question, and this augments that response.
From a simplistic quantitative analysis standpoint, one can look at the spatial resolution of a print, and of the image as captured by the sensor.
With a 16x20 print, it would be safe to assume that the printer is 300 pixels per inch, on a photographic paper. Many printing processes are higher resolution, and one can apply this crude analysis to different resolution printing processes. So at 300 pixels per inch, the 16x20 is:
16inch * 300 pixels/inch = 4800 pixels
20 inch * 300 pixels/inch = 6000 pixels
The approximate MP is 4800pixels * 6000 pixels = 28,800,000 pixel**2 or 28.8 MP
Let's look at the sensor:
DJI
P4P 3:2 is spec'd at 5472x3648 pixels. That is below a 16x20. If it were me, I would just put a border (white?) around the print, and buffer in additional pixels to make the image. 6000 (full size) - 5472 (actual pixels in image) = 528 border pixels. 528 / 2 = 264 pixels per side.
Similarly 4800 (full size) - 3648 (actual image) = 1152 (total border)
1152 / 2 = 576 pixels per side.
In this 3:2 example, with a 300 dpi printer, the rendering of the original image would be:
5472 pixels / 300 pixels/inch = 18.24 inch
3648 pixels / 300 pixels/inch = 12.16 inch.
When one prints an image so that the image has to be scaled, there is an interpolation process to find the new value of a pixel, and this is (usually) an arithmetic average of the percentage of the pixels which effectively map to the rendered image. For example there could be 9 pixels which are mapped in some relationship to one pixel. This tends to change the perception of how "sharp" the image is.
So if I were considering marketing a print taken with a DJI drone with the 3:2 aspect, I would consider offering the print as a nominal 12x18 print, assuming it was printed on a 300 pixel per inch printer, or some integer multiple thereof.
From an aesthetic standpoint, I might consider a product with a 16:9 format, which on the DJI
P4P is 5472 x 3078, which corresponds to a rendered image of about 10x18. Sure you are not using the full 16x20 sheet, but you get a more crisp rendering. If you have a customer who has an eye for detail (hint: myopic customers (nearsighted)) I would go with a crisp rendering, rather than the biggest piece of paper. Of the other hand, if your customer wants it to fit a certain space, or frame, without a mask, then you have to deliver what they want. I suppose if you are concerned about the waste of the border, you could print merge a strip of wallet sized photos on to your final printed image, and cut those off during your framing process, and give them away, or keep them for examples of your work.
So, in summary, generally, you want to have an integer alignment of imaged pixels to rendered pixels for maximal quality. In my earlier answer, I suggested that you try different sizes. Here I have laid out the quantitative work to fit the captured image to the rendered print to (easily) maximize quality.
Stated differently, avoid resampling images.
Perhaps this helps a bit.
Final note: All this assumes that you have a stable platform and good shutter speed to counter motion blur. You will need an ISO setting to minimize noise from the sensor, and you will need an aperture exploit the sweet spot in your lens (don't know what it is for the
P4P, and I should because I have a P3A) but I would guess somewhere around 5.6 or 8 f-stop. Again, you would probably want the sun behind you.
Just one technical note: Higher color gamut can create the impression of greater spatial resolution. Part of this is physics and part of it is human visual system (HVS). Beyond discussion here, however it means that a higher resolution image can provide a deeper color experience for the observer. Similarly, a wider color gamut (think 14 bit dSLR image vs an early 8 bit point and shoot) can create the impression of a higher spatial resolution image. We focus on spatial resolution, but using our sensors in a region where we can exploit the greatest color gamut (or utilizing adaptive mechanisms for gamut extension) will create more unique viewing experiences.
