Downsampling

Trever1t

Been spending a lot of time on here!
Joined
Dec 30, 2010
Messages
9,331
Reaction score
2,722
Location
San Jose, CA
Website
wsgphotography.com
Can others edit my Photos
Photos NOT OK to edit
I've been reading (danger danger) about the huge file size the D800 is going to produce and the benefits of downsampling which I take is decreasing pixle count of the file to a smaller number.

Can someone explain

A) how this increases Dynamic Range?
B) What method using CS5 will produce the best results?


Thank you :D
 
In very basic terms:

Let's say that image noise is more or less random from pixel to pixel, so the noise in neighbouring pixels is uncorrelated. As you add and average uncorrelated data the average tends to zero or settle to a minimum value. Image data is not random and so there is a strong correlation between neighbouring pixels. Adding and averaging correlated data tends to settle to a representative average. Downsampling is adding and averaging data from neighbouring pixels and producing a single value, thereby reducing the effect of uncorrelated data. This does not add (increase dynamic range) new information but lets what data there is come up after the noise is reduced.

I have no idea how CS5 does this, I don't use it.
 
Thanks for the response, what method is used to downsample? Just resizing?
 
Yes typically. Though careful on some algorithms will introduce more noise due to sharpening. It's best to have a play to see how this affects each image.

It's the same phenomenon that allows you to combine two different images to halve the noise, something very often done in astro photography.
 
Megapixels and megabytes aren't the same thing.

The file size, what gets stored on your computer, is about megabytes.

No one method in PsCS5 will give the best results. Image content will determine which image interpolation method works best for that image.

Just before you downsample, be sure and wave goodbye to your original pixels.

The Bicubic interpolations seem to be the best for downsizing.

There are some guidelines. TIFF files are usually significantly bigger than PSD files. The Raw file size won't vary a whole lot.

Your choices in CS5 are:
Nearest neighbor
Bilinear
Bicubic
Bicubic sharper
Bicubic smoother

If you don't already have it, I strongly recommend having at hand for PsCS5 feature/function/capability reference - Adobe Photoshop CS5 for Photographers: A professional image editor's guide to the creative use of Photoshop for the Macintosh and PC
 
As an Amazon Associate we earn from qualifying purchases.
Megapixels and megabytes aren't the same thing.

Trever's question is perfectly valid though. Megapixels are proportional to megabytes as more pixels = more storage space required. If the pixels are pointless (too many for a non-sharp lens, or can be reduced to improve noise performance) then downsampling will result in a lower filesize without any great loss.
 
And I never suggested it wasn't a valid question.

But one wonders - Why buy a 36 MP camera if you're planning to downsample? :scratch:
 
And I never suggested it wasn't a valid question.

But one wonders - Why buy a 36 MP camera if you're planning to downsample? :scratch:

To have 36mp at friendly ISOs, and the ability to get nice 12mp shots at unfriendly ISOs?
 
Exactly. I don't think anyone here is planning to downsample every image. That said I'm sure a large number of D800 owners are also not planning on spending the $zquilliion on the required lens to get decent 36mpxl photos too :)

The point is it's an option. Why would you do it? The same reason astrophotography programs give you the option of bypassing the beyer interpolation of the sensor producing images with 1/4 of the resolution, there is something to be gained in some scenarios that matters more than high pixel count.
 
Right, it's an option available I was just curious as to the mechanics and appreciate all the answers, thank you.
 
Does the camera have the option of shooting smaller files? I don't know Nikons, but the high-MP Canons allow for shooting smaller (about half size I think) raw files. If so, you could switch to the full size only when you think you need it.
 
It does have multiple crop options but file size really isn't a concern, rather I am interested in tweaks that might improve picture quality in particular cases where noise might be present. :)
 
A downsampled image can retain a lot more detail then a native resolution image. The reason being that each pixel on a sensor is only sensitive to one color of the RGB spectrum, the final result is then bayer interpolated--the result is a full color image with less detail then if each pixel was capable of RGB without interpolation.

I would expect about a 25% noticeable sharpness increase in a downsampled image compared to an identically sized native resolution.
 

Most reactions

Back
Top