Possible to adjust ISO sensitivity on each pixel in CMOS image sensors?

Is it possible to design some CMOS sensors which can select a lower
ISO sensitivity value in hot areas and higher ISO on shadow areas? My
pictures tend to have overexposed areas.

It is possible to make different pixels have different sensitivity (Fuji does 
this) but the pixel sensitivity is set at the time of manufacturing and can't be 
altered on a picture-by-picture basis.  So there is no way to change the pixel's 
sensitivity to match your particular picture.

Let me also address your underlying question.  First of all, it is important to 
differentiate between overexposure and too much contrast (or too large a dynamic 
range).  If your problem is really just that parts of the picture are 
overexposed, you could adjust the exposure downward to compensate (underexpose 
slightly).  I know my camera tends to overexpose when set to "normal", so a 
slight decrease using the exposure compensation often helps.

On the other hand, if you end up with pixels that are both 100% white 
(overexposed or "blown out highlights") and 100% black (underexposed or "block 
up shadows"), then the problem is that the scene has too great a contrast range 
for your camera's sensor to handle.  (Use a histogram to see what is happening.) 
Some cameras have adjustable contrast range, so then you can simply decrease the 
contrast to help deal with this.  Also, higher end cameras such as digital SLRs 
tend to have wider and more adjustable dynamic range, so you are more likely to 
capture the scene's entire brightness range in your exposure.  Another technique 
that is sometimes used is to make 2 exposures, one for the highlights and one 
for the shadows, and then combine then in software.  This is obviously limited 
to landscapes and other unchanging subjects and probably requires a tripod. 
Another tip is to use a flash, fill light, or reflector to add light to the 
shadows, thereby decreasing the entire scene's contrast range.  This is 
generally only possible when the shadows are physically located close to the 
camera--e.g. a flash won't help much with a shadowy mountain range 500 yards 
away!  Finally, in landscape work where the sky or sun is visible, a "split 
neutral density filter" is often used to darken one area (usually the sky) while 
leaving the rest unaffected, again decreasing the total contrast range to better 
fit within the camera's limitations.

If none of these are an option for you, you just have to decide if you would 
rather capture the highlights and loose some shadows or vice versa.  That is an 
artistic decision based on the subject matter and what kind of image you are 
trying to achieve.

-- 
Jon Harris
SPAM blocker in place:
Remove 99 (but leave 7) to reply

"Frank" <Francis.invalid@hotmail.com> wrote in message 
news:431bdacb@news.starhub.net.sg...
> Is it possible to design some CMOS sensors which can select a lower
> ISO sensitivity value in hot areas and higher ISO on shadow areas? My
> pictures tend to have overexposed areas. 

It could be the dynamic range that affected my images. I was shooting
portrait
beside the window and intended to have clear background outside the window.
However, on a P&S camera, I got was clear portrait and a white background
with a spot metering on the face.

Is my senario a dynamic range problem or partially overexposure problem? Is
it likely to adjust either ISO or dynamic range or every pixel by software
before
shutter release?






"Jon Harris" <jon99_harris7@hotmail.com> wrote in message
news:TBRSe.12733$Sx4.12076@trnddc06...
> It is possible to make different pixels have different sensitivity (Fuji
does
> this) but the pixel sensitivity is set at the time of manufacturing and
can't be
> altered on a picture-by-picture basis.  So there is no way to change the
pixel's
> sensitivity to match your particular picture.
>
> Let me also address your underlying question.  First of all, it is
important to
> differentiate between overexposure and too much contrast (or too large a
dynamic
> range).  If your problem is really just that parts of the picture are
> overexposed, you could adjust the exposure downward to compensate
(underexpose
> slightly).  I know my camera tends to overexpose when set to "normal", so
a
> slight decrease using the exposure compensation often helps.
>
> On the other hand, if you end up with pixels that are both 100% white
> (overexposed or "blown out highlights") and 100% black (underexposed or
"block
> up shadows"), then the problem is that the scene has too great a contrast
range
> for your camera's sensor to handle.  (Use a histogram to see what is
happening.)
> Some cameras have adjustable contrast range, so then you can simply
decrease the
> contrast to help deal with this.  Also, higher end cameras such as digital
SLRs
> tend to have wider and more adjustable dynamic range, so you are more
likely to
> capture the scene's entire brightness range in your exposure.  Another
technique
> that is sometimes used is to make 2 exposures, one for the highlights and
one
> for the shadows, and then combine then in software.  This is obviously
limited
> to landscapes and other unchanging subjects and probably requires a
tripod.
> Another tip is to use a flash, fill light, or reflector to add light to
the
> shadows, thereby decreasing the entire scene's contrast range.  This is
> generally only possible when the shadows are physically located close to
the
> camera--e.g. a flash won't help much with a shadowy mountain range 500
yards
> away!  Finally, in landscape work where the sky or sun is visible, a
"split
> neutral density filter" is often used to darken one area (usually the sky)
while
> leaving the rest unaffected, again decreasing the total contrast range to
better
> fit within the camera's limitations.
>
> If none of these are an option for you, you just have to decide if you
would
> rather capture the highlights and loose some shadows or vice versa.  That
is an
> artistic decision based on the subject matter and what kind of image you
are
> trying to achieve.
>
> -- 
> Jon Harris
> SPAM blocker in place:
> Remove 99 (but leave 7) to reply
>
> "Frank" <Francis.invalid@hotmail.com> wrote in message
> news:431bdacb@news.starhub.net.sg...
> > Is it possible to design some CMOS sensors which can select a lower
> > ISO sensitivity value in hot areas and higher ISO on shadow areas? My
> > pictures tend to have overexposed areas.
>
>

Frank wrote:
> 
> It could be the dynamic range that affected my images. I was shooting
> portrait
> beside the window and intended to have clear background outside the window.

That's dynamic range that's causing the problem. There is a solution
to this that involves taking two pictures, one exposed correctly for
the portrait (with the window overexposed) and one exposed for the
window (with the portrait underexposed). You can then select each pixel
from the "best" image, either manually or automatically, to make
a composite.

Cheers

mark-r

-- 
"Let's meet the panel. You couldn't ask for four finer comedians -
so that answers your next question..."
 -- Humphrey Lyttleton

Frank wrote:
> It could be the dynamic range that affected my images. I was shooting
> portrait
> beside the window and intended to have clear background outside the window.
> However, on a P&S camera, I got was clear portrait and a white background
> with a spot metering on the face.
> 
> Is my senario a dynamic range problem or partially overexposure problem? Is
> it likely to adjust either ISO or dynamic range or every pixel by software
> before
> shutter release?

With emulsion, and to a smaller extent with most digital sensors, the 
dynamic range of the original recorded data -- negative, pixel values -- 
is greater than what can be reproduced in the display. With film, the 
dynamic range is compressed when printing by "dodging" and "burning in" 
particular areas. With digital images, software can do similar 
operations. Photoshop (costly) and Gimp (free) are two programs that 
allow such manipulation. Of course, once the sensor saturates white or 
black, there are no data left to manipulate.

   ...

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

Mark Robinson wrote:
> Frank wrote:
> 
>>It could be the dynamic range that affected my images. I was shooting
>>portrait
>>beside the window and intended to have clear background outside the window.
> 
> 
> That's dynamic range that's causing the problem. There is a solution
> to this that involves taking two pictures, one exposed correctly for
> the portrait (with the window overexposed) and one exposed for the
> window (with the portrait underexposed). You can then select each pixel
> from the "best" image, either manually or automatically, to make
> a composite.

Sometimes, fill flash is the simplest way.

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

"Frank" <Francis.invalid@hotmail.com> wrote in message 
news:431bf59c$1@news.starhub.net.sg...
> It could be the dynamic range that affected my images. I was shooting
> portrait
> beside the window and intended to have clear background outside the 
> window.
> However, on a P&S camera, I got was clear portrait and a white background
> with a spot metering on the face.
>
> Is my senario a dynamic range problem or partially overexposure problem? 
> Is
> it likely to adjust either ISO or dynamic range or every pixel by software
> before
> shutter release?
>

Hello Frank,

Yes as the others said, you have an DR problem. And Jerry's suggestion is 
the simplest. Set the camera to properly expose the background, and then use 
a fill flash to bring the rest of the subject into proper exposure. You may 
have to set the camera on manual. There are methods for handling very high 
contrast scenes (such as combining shots taken with two different 
exposures), but if compressed to the typical 4/5 F-stop range for printing 
or viewing, they tend to look kind of weird. Usually for portraiture, you 
want soft light so if you have the person by the window, you can add a 
reflector just behind them putting some of the window's light back onto 
their face. This will be less harsh that a simple flash. Or you can add a 
diffuser to the flash or fire it into a reflector. These methods all work to 
reduce the pointance of the light. Light sources with high pointances tend 
to exaggerate skin imperfections.

Clay

Jerry Avins wrote:
> 
> Sometimes, fill flash is the simplest way.

Indeed, but this is comp.dsp after all :-)

Cheers

mark-r

-- 
"Let's meet the panel. You couldn't ask for four finer comedians -
so that answers your next question..."
 -- Humphrey Lyttleton

"Jerry Avins" <jya@ieee.org> wrote in message 
news:bfadnagI9v14yYHeRVn-2Q@rcn.net...
> Frank wrote:
>> It could be the dynamic range that affected my images. I was shooting
>> portrait
>> beside the window and intended to have clear background outside the window.
>> However, on a P&S camera, I got was clear portrait and a white background
>> with a spot metering on the face.
>>
>> Is my senario a dynamic range problem or partially overexposure problem? Is
>> it likely to adjust either ISO or dynamic range or every pixel by software
>> before
>> shutter release?
>
> With emulsion, and to a smaller extent with most digital sensors, the dynamic 
> range of the original recorded data -- negative, pixel values -- 
> is greater than what can be reproduced in the display. With film, the dynamic 
> range is compressed when printing by "dodging" and "burning in" particular 
> areas. With digital images, software can do similar operations. Photoshop 
> (costly) and Gimp (free) are two programs that allow such manipulation. Of 
> course, once the sensor saturates white or black, there are no data left to 
> manipulate.

I agree with you about film, but in my experience, this has not been a problem 
with digital images.  My monitor (when properly adjusted) can display the full 
range of brightness in a standard image file.  And on-line photo printing sites 
handle images that contain the full 0-255 brightness range of JPEG files just 
fine.  In fact, if you do submit a photo with limited contrast, many will 
stretch the tonal range to fill out the dynamic range (the better ones make this 
an option). 

"Frank" <Francis.invalid@hotmail.com> wrote in message 
news:431bf59c$1@news.starhub.net.sg...
> It could be the dynamic range that affected my images. I was shooting
> portrait
> beside the window and intended to have clear background outside the window.
> However, on a P&S camera, I got was clear portrait and a white background
> with a spot metering on the face.

Almost any time you have both the primary light souce (outdoors in this case) 
and the object being lit by that source (face in this case) both in the same 
exposure, you are asking for dynamic range problems.  The light source is 
typically many times brighter than the object reflecting the light so it is 
difficult to capture both completely.  Other common examples are a portrait by 
lamp-light that includes the lamp or an outdoor picture that includes the sun.

> Is my senario a dynamic range problem or partially overexposure problem? Is
> it likely to adjust either ISO or dynamic range or every pixel by software
> before shutter release?

I'll say it again--it is not possible.  Think about what would be required. 
Adjusting "by software" implies that the digital image data is already present. 
But "before shutter release" means that you haven't captured it yet.  So there 
is a chicken and egg problem.  You might think you could make 2 expsosures in 
rapid succession, using the first one to control the pixel sensitivity of the 
second.  But this would require that the image was absolutely constant between 
the 2 exposures and would most likely have problems with sharp edges where it 
transitions from light to dark.  Not to mention that sensors aren't built with 
the capability to adjust individual pixel sensitivty on the fly!

A much simpler solution is to create sensors with wider dynamic range, which is 
increasingly happening.  In the meantime, you are left with the options I listed 
in my first reply and others have expanded on: 1) use fill light or flash, 2) 
use a reflector, 3) take 2 exposures and combine in software or 4) use a split 
neutral density filter.  In your case, #1 or #2 sound like the best options. 
Simply turning on a bright light in the room may be enough to do the trick.

> "Jon Harris" <jon99_harris7@hotmail.com> wrote in message
> news:TBRSe.12733$Sx4.12076@trnddc06...
>> It is possible to make different pixels have different sensitivity (Fuji
> does
>> this) but the pixel sensitivity is set at the time of manufacturing and
> can't be
>> altered on a picture-by-picture basis.  So there is no way to change the
> pixel's
>> sensitivity to match your particular picture.
>>
>> Let me also address your underlying question.  First of all, it is
> important to
>> differentiate between overexposure and too much contrast (or too large a
> dynamic
>> range).  If your problem is really just that parts of the picture are
>> overexposed, you could adjust the exposure downward to compensate
> (underexpose
>> slightly).  I know my camera tends to overexpose when set to "normal", so
> a
>> slight decrease using the exposure compensation often helps.
>>
>> On the other hand, if you end up with pixels that are both 100% white
>> (overexposed or "blown out highlights") and 100% black (underexposed or
> "block
>> up shadows"), then the problem is that the scene has too great a contrast
> range
>> for your camera's sensor to handle.  (Use a histogram to see what is
> happening.)
>> Some cameras have adjustable contrast range, so then you can simply
> decrease the
>> contrast to help deal with this.  Also, higher end cameras such as digital
> SLRs
>> tend to have wider and more adjustable dynamic range, so you are more
> likely to
>> capture the scene's entire brightness range in your exposure.  Another
> technique
>> that is sometimes used is to make 2 exposures, one for the highlights and
> one
>> for the shadows, and then combine then in software.  This is obviously
> limited
>> to landscapes and other unchanging subjects and probably requires a
> tripod.
>> Another tip is to use a flash, fill light, or reflector to add light to
> the
>> shadows, thereby decreasing the entire scene's contrast range.  This is
>> generally only possible when the shadows are physically located close to
> the
>> camera--e.g. a flash won't help much with a shadowy mountain range 500
> yards
>> away!  Finally, in landscape work where the sky or sun is visible, a
> "split
>> neutral density filter" is often used to darken one area (usually the sky)
> while
>> leaving the rest unaffected, again decreasing the total contrast range to
> better
>> fit within the camera's limitations.
>>
>> If none of these are an option for you, you just have to decide if you
> would
>> rather capture the highlights and loose some shadows or vice versa.  That
> is an
>> artistic decision based on the subject matter and what kind of image you
> are
>> trying to achieve.
>>
>> -- 
>> Jon Harris
>> SPAM blocker in place:
>> Remove 99 (but leave 7) to reply
>>
>> "Frank" <Francis.invalid@hotmail.com> wrote in message
>> news:431bdacb@news.starhub.net.sg...
>> > Is it possible to design some CMOS sensors which can select a lower
>> > ISO sensitivity value in hot areas and higher ISO on shadow areas? My
>> > pictures tend to have overexposed areas.