I thought the same, but it seems they display the image 1:1, so no scaling is applied. If that's the case, then I doubt any of the phones will process the image. (Though there might be reasons to do that, which I'm not aware of)
The fact that the author references 600 dpi, then displays the image at 1:1 reveals that he really has no idea what he's dealing with. To further complicate matters, the author's source image is an already anti-aliased bitmap. Re-sizing this image in order to test appropriately on varying screen resolutions will result in anti-aliasing errors. The test images should be rendered at target resolutions from a vector based format like SVG.
This is what the test image looks at zoomed to 400% with no anti-aliasing and no additional processing. This is a pixel-by-pixel resolution of the mess at the center of that image:
When displayed 1:1 on a device, that mess will show up.
DPI is dots per inch. This is an indication of resolution density. When you scale the display of an image on a fixed size screen, you also scale the DPI. That is to say, you could use a 1 million DPI image, and still, when displayed at 1:1 on a smartphone, the DPI of the resulting image is the same (the extra pixels are clipped outside the boundaries of the screen).
Keep in mind that this test pattern should appear like a fractal. No matter how far you zoom in, the pattern will look the same. The points at the center should scale down to near infinite resolution. The resolution limitation is only that of the display device.
A more suitable test that gets closest to the hardware would work something like this:
* Start with an SVG test pattern
* Record the native resolution of each device in pixels
* Render the SVG test pattern with no anti-aliasing (or at least consistent anti-aliasing) at each target resolution
* Display the image on the device using a custom developed application that disables or avoids any internal scaling or image processing capabilities of the software
* Observe the results
If you wanted to test the rendering capabilities of the entire software stack, you should still start with an SVG and use whatever built-in components will render SVG.
Starting from a large bitmap tells us how well the bitmap scaling software works, but you shouldn't zoom to display at 1:1. You should allow the software to resize and see how well it scales.
Each of these is another step removed from the core question asked in the article: Does the pentile screen impact image quality? To answer this question, we must rule out other factors, such as software rendering capabilities.
The final error made is that of attempting to capture the results with a finite resolution capture device (a camera). At a pixel level, your capture would need to greatly exceed the resolution of the target display in order to examine the results at a pixel level or to reproduce the image in a way that objectively represents the perceived result to an audience not present at the time of testing.
