There are good algorithms to rebuild the missing lines without aliasing.

The decision to do it well will be more related to market strategies, give to the low budget people a good hdr image or keep the aliasing free image for expensive cameras.

the 24 megapixel sensor is 6000 x 4000. If they crop the video recording to 5760 x 3240 each exposure will be 5760 x 1620. 5760 is 1920 x 3 and 1620 is 1080 x 1.5, this resolution in a full sensor readout is good maths for a good reconstruction.

if they build a sensor with 33 megapixels so they can do 7680 x 4230 pixels for video recording and each exposure will be 7680 x 2160 enough for 4k hdr or even 8k hdr, but I do not believe they will do this now, maybe in the future.

in doom9 forum they have the QTGMC deinterlace which proves that a aliasing free reconstruction is possible. they just need to be willing to do it, with a powerfull processor in the camera with theys own algorithms, not the QTGMC because it is slow processing.

from the sonyrumors page -

This means both images and recorded video are capable of 15.5 stops of dynamic range!

if true...sony is unstoppable ! this camera will take over the midrange market .

@apefos If it were easy to rebuild missing lines, why is nobody doing it for those pixel-skipping moiré prone cameras?

Anyway, Red with their temporal approach to HDR is showing artifacts in fast movement and I expect the same from this concept, since they also take two different exposures.

Arri's Alexa is doing something similar, and you can see it (albeit in rare situations).

I think this kind of hdr is much better solution compared to s-log2 and v-log because there is no need of post production grading to recover the image, saves time and effort.

I think you confuse two different things.

Log formats exist only because source have much more information than final encoded file. Normal and log approaches are just two different methods to pack this info.

With HDR you, actually, have bigger problem, as you still have 8-bit final file and even wider DR :-)

Proper approach not used now by manufacturers (for consumers and individuals) was to have two independent streams, one non-log normal video and another specially processed and formed video representing all remaining information. So editor can get necessary information in case of grading.

The idea of using two alternate lines for HDR video is not new.

Automotive and Security cameras are using this idea from some time ago because this kind of cameras needs to see shadows and bright areas. Dark garage and bright outdoor, dark tunel and bright road in front are examples.

Also the magic lantern firmware has the dual iso for photo and video which uses this same approach, but they does not have a good reconstruction to avoid aliasing...

I was thinking about how the algorithm could reconstruct the image without aliasing. I started with two ideas: first the simple interpolation to recreate the missing lines using the information from lines above and below, and , to improve the anti-aliasing, using the information from some pixels from each side also. Second the algorithm could get the information from the dark lines and multiply the pixel values using the exposure difference to recreate the missing bright lines, and divide the bright lines using the exposure difference to recreate the missing lines in shadows. 3 fstops difference => shadow lines x 8 = missing bright lines, bright lines / 8 = missing shadows lines. This can give bad results if shutter speed is different for each exposure, but if the difference is in ISO with same shutter speed this can work. A good low noise sensor will be welcome for this. A third option would be to use the two approaches in a more complex algorithm for a better image reconstruction.

If the sensor record 4k video using 3840 x 2160 pixels as the GH4 and A7S are doing, it will be possible to do a good 4K HDR using 3840 x 1080 lines for shadows and 3840 x 1080 lines for bright areas and reconstruct both images and merge. If the sensor uses more pixels in a full readout maybe the reconstruction will be even better.

As I said before: they can do it well, they just need to be willing to do.

I believe the Nikon HoneyComb sensor and Fuji/Panasonic and Samsung organic approaches can be better to avoid aliasing, but there are just patent/initial tests showing it, they are not real yet, not ready to implement, will take more time.

Sounds similar to interlacing. I hate interlacing on progressive displays. There is no way this is going to look better with all of that interpolation going on.

Just give me 10 bit losslessly compressed RAW video output already. I will figure out how to record it onto whatever media I want to store it on.

Some more specs:

• 32mp BSI sensor with 285 PDAF points
• 5-axis IBIS
• 4K video
• Improved EVF
• 14.5fps in photos
• Mic-input

32MP on APS-C?!?! We are in crazy times. I assume this is only possible (with acceptable noise levels) due to the BSI? We just dont need 32 freaking MP, though. Give us the same 24 MP with less noise...

Definitely excited about the other specs, though. We are going to see a hefty price spike with this model.

A mini A7RII?

Recently the rumors about this camera are contradictory; 24MP, 32MP, 4K, not 4K, if 4K, 3 axis, 5 axis, body sealed, not body sealed.

Tomorrow we will know the truth.

@joethepro, has been delayed for february 2016. :(

http://www.sonyalpharumors.com/sr4-a6xxx-announcement-again-delayed-til-february-2016/

Vitaliy, I dont know what think about. The new AXXXX has been announced and delayed several times.

Could it be true that Sony has not beaten the problem of overheating or just interested in the full frame, of course, based on the rumors.