To my mind, the biggest advantage Canon has is the AF system in it's higher end models.... the 5D3 system is certainly nothing to be sneezed at... but think about how it could be improved... Think about recognizing a bird and tracking it through trees! I have a p/s that can recognize and differentiate between a cat and a dog.... surely a DSLR with two higher end processors can do better? The 7D2 might be the camera that takes AF to the next level... and what about linking exposure to focus points? What about using the split pixels to vastly improve DR?
I also hope it gets full 5Dmk3 AFI'd be disappointed if it did.... I'm hoping to see the emergence of a new AF system that will set the standard for years to come...
If the improvements were incremental changes, it would have been out by now. The delays could mean something completely different.
You mean set a standard higher than the 61pt AF system? I think that one will be hard to beat. I've never seen an AF system lock on so fast, consistently, and easily as Canon'a 61pt AF system. What exactly would you improve in it? It certainly isn't broken, and certainly isn't something to be disappointed about if it found it's way into the 7D II.
When it comes to DPAF, that is an entirely different kind of AF for an entirely different purpose. As someone who uses PDAF for the very vast majority of his photography, I don't see DPAF taking over any time soon. While it is certainly cool technology, it's really just the beginning of the leg up that mirrorless needs to BEGIN to compete with what PDAF units currently offer. I think its going to be a few iterations, especially the one where it becomes QPAF, before we see sensor-based AF reaching the level of consistency, speed, and performance of PDAF.
And even then...it would require an EVF to function entirely properly...and EVFs have just as long a way to come before they can really be viable replacements for OVFs...
If the 7D II gets the 61pt AF system, I'd personally be ecstatic!
A 62pt AF system would be vastly superior.
There are a lot of things that could happen and I eagerly await to see which ones they give us.... It should be an interesting release... certainly more than a mode dial that goes all the way around.
When it comes to locking onto and tracking subjects through trees, I do that now with my old 7D and its 19pt AF system. Achieving that is largely a matter of tuning the AF system...getting the right tracking jump rate and using the right point size. The 61pt AF system is far more capable than the 7Ds, and tracking through trees has not generally been a problem for me. Locking on can be tricky, however the 1D X with it's subject identification capabilities using the new metering sensor handles it pretty well. The 1D X also locks onto animal faces...it seems to recognize the general shape of a face, even if it isn't human, and is even capable of doing it in profile.
These kinds of things aren't new, and aren't limited you your P&S. Now, there has always been the metering/AF split between the 1D line and everything else. The advanced integrated meter has always been a "premium" feature. That's a canon thing...they may or may not ever change that, but that is certainly not a limitation of the hardware.
Regarding linking "exposure" to focus points...do you mean link metering to AF points? Again, that's been done, it isn't new technology, it's just one of those things Canon relegates to the 1D line. Otherwise, I don't know what you mean by linking exposure to focus points.
As for split pixels improving DR...I don't believe that is possible. Those pixels receive the same amount of light. Split or a single pixel, the amount of light is the same. Now, assuming you are thinking "read one at ISO 100, the other at ISO 800". That isn't really going to help. It's HALF a pixel. For both ISOs, if you only read half of each pixel at a given ISO, then your halving the signal strength. Noise is the SQRT of the signal, so noise will jump considerably.
For example, at FWC ISO 100 on the 70D you get 26726e-, but at half the pixel, you would only get 13363e-. Your noise at 26726 is 163.5e-, and at 13363 it's 115.6e-. Your signal was halved, but your noise as a ratio of the signal only dropped by 30%. Your ISO 100 "half" image is 41% noisier than if you had used the full pixels. Similarly, at ISO 800 your saturation point is 4055e-. Half that you get 2027.5e-. Noise wise, you have 63.7e- at max saturation, and 45e- with a half pixel. Noise as a ratio of the signal dropped by 30%, but your half-pixel ISO 800 image is once again 41% noisier than if you had used full pixels. It would basically be like using ISO 200 instead of ISO 100, or ISO 1600 instead of ISO 800.
This is just referring to the noise intrinsic to the signal...it has nothing to do with read noise. So, assuming you can use this to reduce the impact of read noise on dynamic range...well, you've considerably increased noise levels by only reading half a pixel...so your probably going to end up with a net...nothing. No gain, no real loss. Maybe you get "cleaner" noise, but dual pixel designs aren't going to be the magic bullet for increasing Canon sensor DR.