Go Back   Steve's Digicams Forums >

LinkBack Thread Tools Search this Thread
Old Nov 23, 2003, 8:35 PM   #1
Senior Member
bradg's Avatar
Join Date: Jun 2003
Posts: 819
Default noise measuring unit??

is there a unit for measuring noise at a set standered iso

bradg is offline   Reply With Quote
Sponsored Links
Old Nov 24, 2003, 11:58 AM   #2
Senior Member
slipe's Avatar
Join Date: Oct 2003
Location: Indian Rocks Beach, FL
Posts: 4,036

Noise is a subjective thing and doesn’t have a unit I know of.
slipe is offline   Reply With Quote
Old Nov 24, 2003, 5:48 PM   #3
Senior Member
voxmagna's Avatar
Join Date: Aug 2002
Posts: 2,162

It's probably a statistical density measurement. That's the easy part. Then you have to relate the objective measure, taking into account size, and frequency of density variations to produce a measure which has some meaningful subjective correlation to impairment or quality.

You could scan image areas which should have uniform density, and analyse the ac frequency components as 'electrical noise' but it still has to relate to what you see in impairment severity, e.g low frequency large area noise in central zone might be subjectively more severe than high frequency random (closely spaced) variations at the outer image frame limits. Then there's correlated noise which might measure like periodic frequency banding. This is really complex academic stuff and you aren't going to measure something like this with a multimeter.

One technique often used is to electronically subtract a reference slide file from a camera recorded image at different ISO levels. The statistical density distribution can probably be got via a fast PC software package analysing the two image files as data. But since JPEG can produce noise like distortion on edges, it probably has to work in uncompressed bitmaps, don't know how you'd spatially align the reference with the captured image though - supered gratings?. VOX
voxmagna is offline   Reply With Quote
Old Nov 24, 2003, 6:41 PM   #4
Senior Member
BillDrew's Avatar
Join Date: Jun 2002
Location: Hay River Township, WI
Posts: 2,512

You can take a picture of a monochrome object (e.g., gray card) and use a statisital measure (standard deviation would be my choice - SD) of the distribution of pixels. They should all have the same value (subject is monochome), but won't because of the noise. Not subjective at all.

Then it gets real messy. Do you measure the SD of the green channel, all channels, only of the L channel in LAB color, ....? At that point most folks throw up their hands and give up. In particular if they are trying to compare noise between different cameras since one might be better for the blue channel, but worse for green.
BillDrew is offline   Reply With Quote

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 On
Pingbacks are On
Refbacks are On

All times are GMT -5. The time now is 8:42 AM.