Digital Imaging Breakthrough

[willie_901]

Eric Fossum recently published a major breakthrough in digital imaging: the Quanta Image Sensor.

This technology uses a digital alternate to analog electronic signal amplification of the CCD /photo-diode signal on the sensor chip itself. It is a fundamental breakthrough in CMOS chip design.

The result is a significant reduction in the analog read-noise level. For digital photographers this technology will eventually deliver significant improvements in shadow region rendering and analog dynamic range. Here is a quote from the Conclusions Section of the peer-reviewed journal article.

"For the first time, photoelectron counting is proven to be feasible in CMOS image sensors without avalanche gain, enabling a new generation of highly sensitive, high resolution image sensors."

During 1993-95 Fossum's research group invented the CMOS sensor... the first fundamental improvement of signal-to-nose ratio for CCD sensor technology. So one can assume digital photographers will benefit from Quanta Image Sensor technologies in 8-10 years.

 

[JP Owens]

Ain't Moore's Law great? At the current technology rate of advancement, in five years we will have cameras with a thousand megapixels, unlimited ISO, and only a fixed lens, because the resolution is so high you can crop in a 100 times without loss.

Can't wait to shoot cat memes with that sucker! 🙂

 

[Spicy]

Ain't Moore's Law great? At the current technology rate of advancement, in five years we will have cameras with a thousand megapixels, unlimited ISO, and only a fixed lens, because the resolution is so high you can crop in a 100 times without loss.

Can't wait to shoot cat memes with that sucker! 🙂

cropping doesn't change perspective (distance relationships in foreground vs background)

 

[mfogiel]

So, you are suggesting, that in 10 years time we actually could get a digital camera that will be capable of nice highlight rendering ? Good news.

 

[codester80]

In ten years there won't be any cameras. The CMOS sensor will be reduced to such a size that it can be implanted in the human eye. We will be able to simply look at something and click a button (or maybe even just blink) to capture an image.

 

[John E Earley]

In 10 years, 2-D images will be old hat.

 

[teddy]

Samples please! 😉

 

[btgc]

In 10 years, 2-D images will be old hat.

Holographers will be urged to try extinct 2D photography experience.

 

[willie_901]

So, you are suggesting, that in 10 years time we actually could get a digital camera that will be capable of nice highlight rendering ? Good news.

They are all capable of good highlight rendering today. The problem is the analog dynamic range of the measurement (i.e. data recorded when the shutter is open) is limited. Even current DSLRs with the highest analog signal-to-noise ratios can only deliver either excellent shadow-region rendering, or highlight region rendering... but not both.

Perhaps you are referring to photon (aka shot) noise? Photon noise is most visible in highlight regions because read noise is essentially absent in regions with high SNRs.

The bad news is photon noise is part of the signal... l which means it can not be minimized during data recording. The good news is quantum mechanics predicts everything we need to know about photon noise characteristics. So selective, post-acquistion filtering (aka noise reduction) will not degrade highlight region details. In fact some brands apply this sort of filtering (usually described as lossy raw compression) to their raw files. This is often criticized by naive pixel peepers. It is trivial to simulate photon noise during rendering.

 

[jloden]

Very cool stuff... and probably a more realistic time frame to expect these improvements to show up in cameras than most people would have.

I have to chuckle a bit when someone announces a sensor tech breakthrough at the research level and people post about holding out for the next camera to be released with the new sensor 😀

Thanks for sharing this.

 

[mod2001]

cropping doesn't change perspective (distance relationships in foreground vs background)

that's why its so interesting if you work with only one lens, zooming by cropping from the same distance, in opposite to 'foot zoom' where you change the perspective.

Yogi