The Panasonic GH4 is barely on the market yet, and while we've seen some comparisons, some have been with cameras that aren't shooting 4K. But what about comparing the $1,700 4K GH4 to the 5K RED EPIC MX (non-DRAGON), which runs well over $20K for a working body? That's what Luke Neumann did, and he threw in the Canon 5D Mark III shooting Magic Lantern RAW for good measure, upscaled to 4K to keep everything consistent. Can a cheap camera really look sharper than a pro cine camera?
[Update 2] Luke also did an additional test with the GH4 and EPIC which you can download here. As I mentioned below, the Panasonic is doing a lot of processing internally, even if you turn down sharpness, which he did with the second test. The EPIC, on the other hand, does essentially nothing to the image, meaning there is no additional sharpening going on in the hardware. You can always add sharpening in post to the EPIC, which should bring it closer to the GH4, but the images still speak for themselves.
Nikkor Ai-s 50mm f 1.2 @ f 5.6
Shutter at 180
CineLikeD: Contrast and Saturation to -5, Sharpness to -2
Nikkor Ai-s 50mm f 1.2 @ f 5.6
RedLogFilm: Contrast and white balance adjusted to match GH4 as close as possible.
This is a still from the new test (click for larger), which has been blown up to 400%:
[Update] Luke also shot this video with only the GH4:
What's Going on Here?
There is no problem with the focus of the lens (it's also stopped down where it should perform better), and Luke has actually uploaded samples so you can check them out without YouTube compression. First things first, Luke owns that EPIC, and I also own a RED, and we both own cameras of all brands -- basically we don't care what comes out on top in the end.
There is a huge difference between resolved detailed and perceived sharpness. Images can look very sharp because of edge enhancement (or high contrast), but may not actually be recording that much fine detail. Essentially there is an infinite amount of detail in the visible world, and how well a camera system reproduces this fine detail up to their pixel count is talked about in terms of resolved detail. A number of factors can affect resolved detail, including the lens, and the filter over the sensor (optical low-pass filter) that prevents aliasing or false details from showing up in the image as a rainbow pattern.
Nearly all color CMOS camera sensors (essentially most cameras you're using), have what's called a Bayer pattern:
If you look at a straight Bayer pattern image it looks terrible, and so in processing this pattern must be interpolated with surrounding pixels to make an image that looks like the scene you were shooting. This means that the image coming into the camera can't be translated 1:1 with your final image, and thus you're losing some resolved detail having to Debayer the image and interpolate. The best Debayering algorithms can reproduce about 75-80% of the detail of whatever your initial resolution is, meaning a 4K color sensor can never actually resolve 4K detail, but 5K should come closer. This is also why 4K downscaled to 1080p looks good, because you're sampling from many more pixels, and the whole system is already resolving well over 1080p.
Yes, the Panasonic GH4 Looks Sharper
Cameras that shoot RAW video, like the RED EPIC and 5D3 Magic Lantern RAW, do minimal processing internally to the image. Some has to be done to actually give you a clean image out of the box and remove fixed-pattern noise, but for the most part RAW is left to post. The Panasonic GH4, on the other hand, is a highly compressed camera, and it aims to give you the best possible image right to the card in the camera. This means it's doing a lot more processing, which includes internal sharpening.
Is it actually resolving more detail at 4K compared to the RED's 5K downsampled to 4K? It may very well be, and regardless of what you can do in post to an image, the vast majority like something that looks nice right out of the box. Could you sharpen the 5D Mark III and RED to look more like the Panasonic? Possibly, but in the end, for many people, the camera that costs the least, looks "the best," and requires the least amount of work to get to a satisfactory image might be the one they favor, especially if it's stuff going to the web where there may only be the slightest difference between a $2,000 camera and a $100,000 camera thanks to all sorts of factors like compression and screen resolution.
In the end, there are many factors that affect image quality, and resolved detail/sharpness is just one of them. There are plenty of others like color, dynamic range, motion rendering, etc. Each of these cameras has their own unique look, and they also have their own usability differences, which are often more important than the image, and are why one camera might be right for one job while another is not.
The test above shows that Panasonic is pushing as much detail as possible through this new 4K sensor, and with minimal correction, it stands up next to a camera that costs 10 times as much. Granted, the MX sensor is quite a few years old at this point, and DRAGON likely performs better, but we have now reached a point where you can get 80-90% of the overall quality of those high-end cameras for a fraction of the cost.