The Fundamentals of Downsampling in Photography

ememovic (67)in #steemstem • 8 years ago (edited)

Hello guys,

I thought it would be fun to expound about Downsampling in photography and hopefully, it'll be pretty useful for any of you who are interested in making images or even videos.

What’s Downsampling?

You may have already heard of this term before, essentially, what you're doing with Downsampling is taking an image at a higher resolution than you expect to use finally as the final image and you're shrinking that high-resolution image with lots of information to a smaller one that is smaller but higher quality, so you're getting a higher quality image that's smaller as opposed to a larger image overall.

_[1]

Why do you need Downsampling and why is it important?

Most cameras today, pretty much all of them actually use what are called “Bayer Sensors” and that just means that for every pixel, one captures red, one captures blue and two capture green but all you need to know is to actually capture one pixels worth, you need a red, green and blue. We add the extra green one in there to make it look better to our eyes which are more sensitive to green, but you can think of it as needing three real physical pixels to capture one pixel that is going to be used in displaying it back on a screen.

_[2]

_{[The Bayer arrangement of color filters on the pixel array of an image sensor -Wikipedia]}

If you look at a specification that says 1920 x 1080, it's really not actually true 1080. That's because it has to be interpolated because the part of the pixel array that's capturing one part of scene is only blue, green or red, you're not getting all the information, the rest has to be built back in, so the effective resolution is a lot lower than what they're saying.

_[3]

375px-Bayer_pattern_on_sensor_profile.svg.png

_{[Profile/cross-section of sensor -Wikipedia]}

They use some algorithms that interpolate and judging by what these different pixels captured, one is probably going to be more of a orangish color so you're gonna have to get more information to get a real image. That's where Downsampling comes in, if you capture a 4k image to Downsample into a 1080 image, that's gonna mean that you're getting way more effective pixels and even if you consider that they're bayer pixels, you're still getting enough to have true 1080 pixel. This is the reason why a lot of cameras claim to have a 1080 pixel and then when you look at another camera and it looks way sharper even though it's still 1080.

_[4]

4k-downsampled-vs-1080p.thumb.jpg.33683754c619fd423b69c6d3a5860243.jpg

_{[4K image downsampled to 1080p -Eoshd.com]}

What's the difference?

One camera is gonna be capturing the original image that’s 4k, it will shrink it down and it's going to look much sharper because it has much more original information to deal with, it's going to look truer and it doesn't have to interpolate any information, it's capturing how it actually is and there's no guesswork.

Software does have a lot to do with it, if you captured an image ten years ago, it would be better looking if you imported it into a raw editor today than you did ten years ago because the algorithms for adapting that information is gonna look better. But in any case, it's gonna always look better if you're using CMOS sensor to Downsample, then it would be to capture at the native resolution.

If you ever see a camera such as the C100 that displays “Effective Pixels”, it's simply talking about the actual physical number of pixels which are bayer pixels, hopefully that made sense.

I kind of had a hard time wrapping my head around myself, if you guys have any questions or if you want me to explain a little bit more, I can try to do that in the comments. If you found this article useful, let me know what you think and don’t forget to smash that upvote button.