A disclaimer. I am not an expert in neurology, though I have some knowledge of it. This post is a philosophical exploration of a mechanism that I believe, with some modification, could work. From this, however, it is encouraged to be inspired, be it as wrong or a right as may be.
John is looking at a lilac. The light is entering his eyes, where it hits his photoreceptors, and goes through the familiar dance. First, already in the optic nerve, some clever computation is being done. Points of difference are enhanced through a dampening effect, applied from all neurons to it's neighbor based on the strength of stimuli. Strong stimuli will dampen even more strongly those neighbouring neurons that receive less, and so a boarder is set between foreground and background. This helps, but, yet the brain has not made that distinction.
Instead, features must be bounced backwards, into the brain. Those neurons which before has answered to the lilac colour, now again fires, more vehemently than all others. Similar stimuli, all the way from the front are enhanced through the mechanism "neurons that fire together wire together," and now the lilac colours of the eye, and the lilacs of the brain are singing together.
The astute observer of colour has many lilac cells, each responding more or less to the shade. But there are also other factors like shape. Shapes are shown on the eye in certain patterns, and those signals, like a sound wave pattern is routed through that shape sending neuron as far back as differentiation will go. And so on for the distribution of shapes.
You can see the signal rushing to the left. Above, the lilac, then the grape like shape, and at the bottom distribution. Then the return signal comes. It bounces back according to the distance and all the patterns of the neurons firing are exactly the same as they would ever be, only, they come in in a slightly new way. Because John has never seen any lilac flower just that way before, from that angle. But this is not enough.
From this ecco we have a singular waveform that is a like a picture of the features collected. Now the signal needs to hit a neuron on the return side, which will fire according to the number of features that correspond to an earlier impression of lilacs. It will fire weakly for few, and strongly for many. I envision that there must be some division of labour here too, because it would not make sense for the whole brain to review every echo of impressions all the time. But exactly how such differentiation happens, I have no inkling.
What ought to happen, is that the signal of only this neuron firing on the collection of features, reroutes it's "lilac" signal to the conscious part of the brain.
Not everything in the visual field is parsed, only that which falls within our focal vision, abut an inch squared at a meters distance.
Oft seen collection of impressions will get their own neurons, which can later receive names through association.
All in all this seems like a reasonable mechanism. I can't wait to ask someone if this holds up.