Hogloff wrote in post #15478465
Maybe for you, but I would rather have scientific repeatable tests that are duplicated on different cameras so we can compare apples to apples. Taking a picture of a cat is not reproducible, repeatable between cameras or in fact with the same camera. Also, sharpness is but just one attribute of a camera that needs to be tested and in fact the sharpness of your cat is more dependent on the lens and not the camera.
Most of the examples that people post from their cameras are fairly meaningless, for the purpose of demonstrating their camera's abilities and limits.
You see downsampled images presented as "sharp", but really, anyone who can't get a sharp result for a web-size image needs to go to ImageProcessing 101, even from a joke camera. Or, someone does a tiny image downsized with "Nearest Neighbor" that is clearly DISTORTED with extreme aliasing, to which people reply, "WOW! What detail and clarity!!!".
If you talk about your experience with AF on a camera, people will ask for an example, never once stopping to think that an example of a sharp result tells you absolutely nothing about all the shots that miss and are never shown; proof of AF accuracy or speed would require 100% crops of a long series of images taken with none of them omitted.
Then, there's the "here's my FF camera at ISO 3200" where they show us a 450x300 image, which is not challenging even for a 1/2.3"-sensor compact these days. ISOs in the hundreds of thousands are where we should expect to see major differences at that size. I cringe when I see how easily people are impressed with technical abilities of cameras from samples that come nowhere near the real issues of the limits of current imaging products.
Of course, DxO doesn't show the whole picture, even for the RAW data. It ignores line noises, which are much more image-interfering than their contribution to a monolithic standard deviation would suggest. Being that Canon has more line noise than Sony or Nikon sensors, DxO actually handicaps Canon performance, giving Canons better scores than they actually deserve in many cases, in read-noise-based measurements, especially for the 5D2 which has pretty high levels of line noise for a relatively recent camera.