Re: lightness metric for images
- From: "aruzinsky" <aruzinsky@xxxxxxxxxxxxxxxxxxxx>
- Date: Tue, 29 Aug 2006 11:52:57 -0400
"I convert the image to the L*a*b*, apply a Gaussian filter on the L layer,
compute the average of the filterred L value on the image, and finally use
it as the metric. Does it work?"
Even if I knew your exact purpose, your question is best answered by
experimentation. So what if you reinvent the wheel? I can tell you that
Gaussian blur won't change the overall average.
What most people do to manually adjust contrast and brightness is to look
at the histogram and
1. Subtract a constant so the left part of the histogram is at zero.
followed by
2. Multiply by a constant so the right part is at or slightly above 255.
This is done while inspecting for "blown highlights."
Anomalies such as dead and hot pixels are excluded from consideration
because they do not show well in the histogram graph. In an automated
process, dead and hot pixels can be removed by median filtering, but not
by Gaussian filtering.
Does this help?
.
- References:
- lightness metric for images
- From: VijaKhara
- Re: lightness metric for images
- From: VijaKhara
- lightness metric for images
- Prev by Date: Re: Image processing prototype software
- Next by Date: Re: image pattern matching problem
- Previous by thread: Re: lightness metric for images
- Next by thread: Re: lightness metric for images
- Index(es):
Relevant Pages
|