MIZZI Computer Software

Digicam noise filtering example

   
Here we show an example how noise filtering is working: The following shot was taken with my KODAK DX6440 4MP camera in a room with 50% daylight and 50% tungsten light. The aperture was f/4.8, exposure time 1/30 sec, focal length 90mm (35mm equiv.), ISO 200, exposure bias -0.5. The shot is to dark but the noise is still acceptable.
f/4.8,
1/30 sec,
ISO 200
click to enlarge it

Let us correct the image: first correct the the day/tungsten light then make it lighter by at factor of 2. The image corresponds now to ISO 400 or higher (due to the tungsten correction). You can now see the noise if enlarging the image (click to the image). The noise becomes unacceptable for bigger prints.
f/4.8,
1/30 sec,
ISO 400 equiv.
click to enlarge it

Let us crop the eyes of the girl and show the image without any compression. We see the unacceptable noise ...

now we apply our noise filter ....

Can you believe the result? The noise is almost removed. The hairs at the right are sharp and crispy as before and no freckle has been removed.

For a better comparison we have used an other crop region showing more of the image. This region corresponds to a 1.7 MP shot. It is compressed by 2 and if your screen has 70-100 dpi you should see images of the size of a medium sized print (15x20cm).

and apply the noise filters ...


You may now think: all this is working fith with a Kodak DX6440 .... Then look to the following image from a Panasonic FZ20 at ISO 400. First a cropped region from the original image without any noise filtering is shown. Click to see the uncropped image.

and now the noise filtered one ..... (click to see the uncropped image)

A nearly noise free picture good for every print up to A5.
Let us compare it to the image the ISO 100 image (taken at the same time).

The ISO 400 image has nearly the same noise but is still a little bit more blurry than the ISO 100 one.
However we can also apply the filters to the ISO 100 image (some moderate filter) .....
... Yes indeed, if the original is better the filtered one is also better.