So I'm sitting here thinking about a new method for increasing dynamic range.

[fuzzybabybunny]

So I'm thinking about ways to increase dynamic range. Currently, dynamic range is very hard to increase with sensors due to physical limitations of the photo sites. Basically, it is very hard to physically manufacture higher dynamic range capabilities into photo sites.

So I was thinking about something that might be described as spatially variable ISO (SV ISO).

You have a classic high dynamic range scene:

http://fuzzybabybunny.smugmug....photos/111983409-L.jpg

If you were to shoot this with only one shot with current technology, you would get:

1. Properly exposed sky but severely underexposed crater and the back of the arch.

or

2. Properly exposed crater and arch but severely overexposed sky.

or

3. Crater, arch, and sky that are all not very well exposed.

What if there was a system in a camera that could analyze each section of the scene, determine the brightness of each section, and spatially adjust the ISO so as to properly expose each part of the picture?

Ex.
In just one shot, the sensor would do this:
http://fuzzybabybunny.smugmug....photos/195201844-O.jpg

The camera would increase or decrease the ISO at specific parts of the sensor to match the scene. So areas of the sensor that would register the bright sun would be set at a low ISO to prevent overexposure of that area, and areas of the sensor that would register dark areas of the scene would gain up to higher ISOs to prevent underexposure, and everything in between would be set accordingly. Obviously this would mean that each independent photo site would be able to be assigned a different ISO, or signal gain, which is something that I think may be possible to build into sensors. It could even be fluid, like the center of the bright sun would be shot at the lowest ISO possible (ex. ISO100), and as you get progressively further from the sun ISOs would increase to 110, 120, 135, 145, etc.

Just throwing this idea out there. I don't know why, but I'm getting this weird feeling of deja vu, like I've explained this idea before...

 

[magomago]

sounds interesting...but are you assigning each site with the ability to have a different ISO, or the screen itself is cut into preset chunks (lets say 16 total squares) that could adjust? I'm thinking it sounds ilke the former. From an imaging standpoint I think the main concern is having the noise make the picture look un-natural, and from a hardware standpoint how much more silicon will be needed to control each site.

It definitely sounds like what it will probably move towards in the future. If not that idea, then something like give an mode where it takes 10 shots at once at different exposure compensation levels and then preform a type of HDR in hardware to generate a single image.

 

[fuzzybabybunny]

Originally posted by: magomago
sounds interesting...but are you assigning each site with the ability to have a different ISO, or the screen itself is cut into preset chunks (lets say 16 total squares) that could adjust? I'm thinking it sounds ilke the former. From an imaging standpoint I think the main concern is having the noise make the picture look un-natural, and from a hardware standpoint how much more silicon will be needed to control each site.

It definitely sounds like what it will probably move towards in the future. If not that idea, then something like give an mode where it takes 10 shots at once at different exposure compensation levels and then preform a type of HDR in hardware to generate a single image.

Yup, I'm assigning each site with the ability to have a different ISO. Having preset chunks would just be too blunt.

With low noise sensors I don't think this will be too much of a problem. Add in some noise reduction software for the high noise areas and it should look ok.

The problem is that you are essentially getting a picture that has numerous different exposures within the same picture. This can result in the picture looking unnatural if these exposures aren't managed well. For example, imagine the shadow parts of the picture being TOO light if too high of an ISO is applied. I think that this potential problem could definitely be avoided or lessened by some good intelligence programmed into the camera.

As for the hardware, it would definitely be pretty tough to somehow produce gain controllers for each and every photo site. I still think this would be more feasible than trying to gain dynamic range the old fashioned way, that is, trying to make the photo "buckets" deeper or trying to increase the photo site's surface area, the latter being almost totally impossible with current high resolution sensors.

The feature to bracket 5+ shots and compile an HDR in-camera would be good, but would still just be a band aid. Obviously this would not work for moving objects (cars, ocean waves, rustling leaves on a tree), unsteady hands, etc.

 

[ElFenix]

changing the amplifier is one way, a way i was thinking of was by doing electronic shutter for each photosite. that would give proper exposure for each part without the gain in noise. implementation might be difficult, however. you'd need to have spot meter with a highlight/shadow key on the camera (was it the E-1 that had the feature?), plus some sort of tracking using an image sensor so that the camera could remember where in the scene the spots are relative to each other.


i really need to talk to my patent lawyer friends. i've also got one for how to solve the large image sensor/telecentricity problem for compact cameras.

 

[fuzzybabybunny]

Originally posted by: ElFenix
changing the amplifier is one way, a way i was thinking of was by doing electronic shutter for each photosite. that would give proper exposure for each part without the gain in noise. implementation might be difficult, however. you'd need to have spot meter with a highlight/shadow key on the camera (was it the E-1 that had the feature?), plus some sort of tracking using an image sensor so that the camera could remember where in the scene the spots are relative to each other.

i really need to talk to my patent lawyer friends. i've also got one for how to solve the large image sensor/telecentricity problem for compact cameras.

That's also a good way to do it. But obviously, just like my idea, it still has its downsides, namely the quickest overall shutter speed would be limited by the speed of the slowest individual shutter. Not too big of a drawback though since it doesn't impact image quality. And I agree that metering all the individual points would be pretty crazy to implement. Not to mention I don't know how feasible it would be to have such a fine level of individual control for each and every photo site.

Not too sure what you mean by the large image sensor/telecentricity problem for compact cameras You mean if we install a large image sensor in a compact camera we would have telecentricity problems?