It's meaningless and actually misleading marketing speak:
A direct image sensor is an image sensor that directly captures red, green, and blue light at each point in an image during a single exposure.
The sensor doesn't capture blue, green and red neatly in each layer as applied by the diagram, the top layer is sensitive to red, green and blue (it's effectively unfiltered), the middle layer is sensitive mainly to red and green (but is still partially sensitive to blue), and the bottom layer is mostly sensitive to red light (but is still partially sensitive to green and to a lesser extent also blue light).
If you were being very generous you could call it a direct white, yellow, red sensor, but seeing as nobody wants a camera that shoots WYR images it has to be converted into RGB. This requires a lot of complex calculations to be applied to try to separate the colours. I say "try to" because there are some cases when it can't, leading to errors.
In addition to this, due to depth diffusion varying degrees of sharpening and other processing are required for each layer. This is one of the reasons processing Foveon RAW images is very difficult, and support from major RAW editing software is completely missing (there's only two converters I know apart from Sigma's own who have even attempted it).
So the idea there's anything "direct" about their approach is I'm afraid completely false.