[opendtv] Re: Image quality
- From: Tom Barry <trbarry@xxxxxxxxxxx>
- To: opendtv@xxxxxxxxxxxxx
- Date: Sun, 13 Nov 2005 11:28:56 -0500
Still, it seems it would be a mathematically simpler problem to just
deal with the distortion caused by limited optics. I guess the
convolution math there is fixed and known since you described it it a
previous post. Though I would not off hand know how to go about fixing
it.
It almost seems as if you could auto-tune something like this just by
taking pictures of known test patterns and evolving some deconvolution
parameters using various AI techniques like neural networks or genetic
algorithms.
Trying to deal with sources of unpredictable noise or motion seems a lot
harder.
- Tom
Mark Schubin wrote:
> There has been extraordinary work on image deconvolution over the past
> four decades, starting no later than Tom Stockham's work at the
> University of Utah in the 1960s. Even then, he was able to deconvolve
> blurring caused by camera motion (long exposure in a moving vehicle) and
> incorrect focus. Siemens has demonstrated faster deconvolution much
> more recently.
>
> Unfortunately, as I found when I worked with Stockham on a contrast
> compressor in the '70s, while this stuff works great if you need to read
> a secret message, when it comes to entertainment-quality moving-image
> sequences, it (at least as far as I know to date, including
> conversations with Siemens earlier this year) leaves much to be desired.
>
> On the original point about the human visual system, it has the huge
> advantage of having to deal with only what the observer wants to see. I
> think it was Donald Fink who said human vision could be satisfied with
> about 50 bits/second -- assuming it was the right 50 bits.
>
> Consider the relatively simple case of dynamic resolution. It's
> terrible in the human visual system (i.e., we get very little spatial
> resolution in motion). Some researchers have attempted to use that to
> trim HDTV data rates (high spatial resolution at low temporal rates and
> low spatial resolution at high temporal rates). Unfortunately, humans
> can track moving objects, at which point they become static and need
> full resolution.
>
> As you read this message, you need the word you're reading right now to
> be legible, but the others could all be blurry with no harm to your
> reading, as long as there was a sufficiently good eye-tracking system to
> sharpen what you needed as you moved back and forth in the text.
> Lacking that, your computer has a much more detailed screen that the
> human visual system demands.
>
> Ah, well. Maybe someday....
>
> TTFN,
> Mark
>
>
> Tom Barry wrote:
>
>
>>If you know the math of the distortion it is sometimes possible to apply
>>some deconvolution scheme to clear up a image. From Wikipedia
>><http://en.wikipedia.org/wiki/Deconvolution>:
>>
>>---quote---
>>Optics
>>
>>In optics, the term "deconvolution" is specifically used to refer to the
>>process of reversing the optical distortion that takes place in a
>>microscope, telescope, or other optical instrument, thus creating
>>clearer images. It is usually done in the digital domain by a software
>>algorithm, as part of a suite of microscope image processing techniques.
>>Deconvolution is also practical to sharpen images that suffer from fast
>>motion or jiggles during capturing. The early Hubble Telescope images
>>could be sharpened by deconvolution.
>>---/quote---
>>
>>I wonder if the human visual system does any such thing? Can better
>>cameras do this?
>>
>>- Tom
>>
>>
>>
>>
>
>
>
> ----------------------------------------------------------------------
> You can UNSUBSCRIBE from the OpenDTV list in two ways:
>
> - Using the UNSUBSCRIBE command in your user configuration settings at
> FreeLists.org
>
> - By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word
> unsubscribe in the subject line.
>
>
----------------------------------------------------------------------
You can UNSUBSCRIBE from the OpenDTV list in two ways:
- Using the UNSUBSCRIBE command in your user configuration settings at
FreeLists.org
- By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word
unsubscribe in the subject line.
Other related posts:
