Go Back   Steve's Digicams Forums > Digicam Help > Newbie Help

Reply
 
Thread Tools Search this Thread
Old Sep 11, 2003, 8:07 PM   #1
RSD
Member
 
Join Date: Sep 2003
Posts: 64
Default Digital Zoom

i dont really know what digital zoom is
at first i thought it was the same thing as manual zoom just digitally but the guy at the stores said that's not it
so what i think know is that digizoom is how much you can zoom in using the lcd or computer after you've taken the pic?
am i right or way off?
RSD is offline   Reply With Quote
Sponsored Links
Old Sep 11, 2003, 10:16 PM   #2
Moderator
 
Join Date: Jun 2002
Posts: 1,139
Default

Digital zoom is an electronic crop and interpolation back to full sized image.

Truthfully, only a few digicams have a digital zoom which is usable for good photography, in my opinion.

The Sony DSC-F707 and DSC-F717 have 2x digital zoom on top of the 5x optical zoom. Though you loose considerable resolution, the higher the resolution you begin with, the better the results will be with digital zoom. On these two Sony's it's actually usable. On most of my digicams which have digital zoom it's not worth using.

Lin
Lin Evans is offline   Reply With Quote
Old Sep 11, 2003, 10:57 PM   #3
RSD
Member
 
Join Date: Sep 2003
Posts: 64
Default

i know that crop means to make it smaller but what is interpolation?
RSD is offline   Reply With Quote
Old Sep 11, 2003, 11:41 PM   #4
Moderator
 
Join Date: Jun 2002
Posts: 1,139
Default

When a digital image is cropped, as you know it represents a smaller portion of the whole. This smaller image then has fewer pixels comprising the whole and the dimensions are reduced by the amount of the crop.

To get the image back to the full file size as the original capture, software (in the case of the digicam "firmware") known as interpolation is used to add pixels until the actual pixel count in this case is identical to the original before the crop.

The way this occurs is that the firmware examines adjacent pixels and creates either an identical value or an intermediate value depending on the actual numerical value of the pixels on all sides of the one being tested. This procedure is repeated throughout the file. Once these intermediate or same values are determined the firmware actually creates additional pixels until the original file size is restored.

How well this works depends entirely on the amount of detail in the cropped portion. On and image representing a relatively narrow field of view such as a head and shoulders portrait where the head is cropped, etc., it can work very well and little degredation of the image quality results. On images with much fine detail such as wide angle landscapes, there is usually insufficient detail for the interpolation algorithms (formula) to do the job properly and poor images usually result.

Best regards,

Lin
Lin Evans is offline   Reply With Quote
Old Sep 12, 2003, 9:03 AM   #5
Senior Member
 
Join Date: Aug 2002
Posts: 2,162
Default

You could say it's also like taking the photo, printing it and looking at a part of the photo with a magnifying glass. The problem is you may have used lots of Mpixels to capture the original image, but now you are looking through a 'window' at the magnified image, the resolution and sharpness will deteriate - that's what your output print now is and you can never go back to improve it.
voxmagna is offline   Reply With Quote
Old Sep 12, 2003, 1:08 PM   #6
RSD
Member
 
Join Date: Sep 2003
Posts: 64
Default

ok
so cropping doesnt deteriots(sp?) quality but interpolation does
i get it
RSD is offline   Reply With Quote
Old Sep 12, 2003, 1:56 PM   #7
Moderator
 
Join Date: Jun 2002
Posts: 1,139
Default

Quote:
so cropping doesnt deteriots(sp?) quality but interpolation does
No - not necessarily. Interpolation simply uses available data to create new data. If the available data is good (real) then the data created is good. If the available data is garbage, then you get larger garbage.

What is "good data" to the interpolation algorithm? It's data which accurately represents the image. Let me explain like this. You look at a picture in a newspaper and recognize a small picture of a friend.

You take a magnifying glass or microscope and look closely at the face. Do you actually see a nose or eyes? Probably not. Instead you see only a few "pixels" which represent dark and light and position. These are commonly referred to as "marker pixels." Because the human brain can "create" it's own reference from past memories and a few marker pixels we "recognize" our friend's picture. It's like accepting a few brush strokes as "pine needles" on a tree in an oil painting. As long as we don't have to see it under magnification we are quite willing and able to accept these few brush strokes as "good data". But when we are forced to look at it up close under magnification, the illusion dissolves and we see it for what it is - brush stroke on canvas.

So it is with a photo print. But when the interpolation algorithm looks at the pixels it sees what is "really" there. It then duplicates faithfully what it finds and greatly enlarges the image so that if it finds the digital equivalence of "brush strokes," that's what we get in the enlargement. If it finds detail represented by sufficient resolution in the capture, then the print looks just like it should only larger.

So what determines whether the "data" is good or bad? Four factors: resolution, distance from lens to subject, field of view and the amount of fine detail in the frame. The larger the field of view (wide angle) and more detail (distance and/or small objects) the more resolution is required to properly define the extent or "boundaries" of this detail.

So back to my original explanation - head and shoulders portrait; not a lot of detail and crops and interpolation works well. Distant landscape at wide angle and lots of detail captured with insufficient resolution? Not a good candidate for interpolation.

Lin
Lin Evans is offline   Reply With Quote
Old Sep 12, 2003, 4:08 PM   #8
Senior Member
 
Join Date: Aug 2002
Posts: 2,162
Default

Quote:
So back to my original explanation - head and shoulders portrait; not a lot of detail and crops and interpolation works well. Distant landscape at wide angle and lots of detail captured with insufficient resolution? Not a good candidate for interpolation.
Yes, your last point Lin is very relevant with digicams. The close up and macro shots we take seem to be impressive, more so with images with a high 'recognition factor' like human faces and recognisable subjects, but it's when you look at wide landscape shots when printed, that resolution from capture to print often shows as softness at the limit of detail and a picture can lose its 'sparkle'. With film we might have improved this by better lenses because I suspect film media was not the limiting factor. But current digicams have in built optical bandwidth filters and limited pixel density, so I'm not so sure that a better lens can improve on resolution without more Mpix to match.
voxmagna is offline   Reply With Quote
Old Sep 17, 2003, 7:41 AM   #9
Member
 
Join Date: Sep 2003
Posts: 46
Default

As this is the newbie help forum I do not want to be out of line. I liked Lin's excelent explanation which is I guess the artist or intuitive explanation of what is going on.

Now I do not know what your background is (or that of other people interested in this topic) but if you are a bit more technically inclined or intrigued by science, read on. If not, stop now because you will only be more confused (or worse, very bored).

If you want to know more about the scientific principles I suggest you find some book on lenses at the local library. A lens is an optical (lowpass) filter. I could start explaining what that means but it may not interest you at all, so if it sparkles any kind of interest please try and find a physics book on lenses or signal theory (in this sense I actually mean a lens system as the transfer function is not only determined by the lens itself but also by the aperture and the focus, when the system is out of focus only very low frequencies will pass through the system, hence very blurry pictures). So in general the high frequencies represent the high level detail information that Lin is talking about, but also the high frequencies are involved when we observe sharp edges in an image. Less sharp edges means the signal contains less high frequencies.

Now if you got that far, then you need to consider what digitization does to the signal (image) that is left after going through the lens. This will bring you into the realm of Nyquist frequencies etc. Also on this there is lots of literature. Understand that there are two discretization steps being done in a digital camera: 1) the image is spatially discretized, resulting in pixels, 2) the amplitude of the signal (the signal strength) is discretized as well, resulting in a number representing the light intensity typically 8 (a number between 0 and 255) or 12 bits (0 to 4095) per color channel (RGB).

The frequency content of your original photographic subject, in combination with these effects (optical filtering and digital sampling) will determine how succesful interpolation may be and which interpolation approach (there are many) will lead to acceptable results.

Two more things need to be considered: image compression and image sharpening.

On compression: the digital image is often further processed by compression algorithms that may or may not change the underlying data, in order to reduce the amount of data that needs to be stored. Of course when data is being removed (lossy compression, or reduction) this will influence the image quality and thus all the things that I mentioned previously (image frequency content) and it will influence the effect of interpolation as well.

Image sharpening: most digital cams (if not all) do some post processing after the image has been captured, basically it tries to "compensate" for loosing some of the higher frequencies by the effects I have described. This processing will also influence the results you can get by interpolation. Also it exploits some of the sensitivities of the human vision system in order to fool you to think the image actually contains more details (is "sharper") than it actually is.

Hopefully this was of any use to someone (not that I am telling anything new here), if so let me know also if you want some more detailed information on transfer functions and the like

Finally do not forget that everything you see also passes an optical system (your eyes) and is post processed (by a cpu called your brain), so what comprises a good photo has at least survived two optical systems, two processing units, at least one analog to digital conversion (in the camera) and one digital to analog conversion(when the photo is displayed) and at this time only god knows exactly what kind of conversion takes place in your brain we are still working on understanding that part.
BigMac is offline   Reply With Quote
 
Reply


Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off



All times are GMT -5. The time now is 12:38 PM.