Edge enhancement comes from lateral inhibition, not a change in physical luminance

What it is

The subjective boost seen at a Mach-band edge traces back to lateral inhibition in visual-system neurons and the center-surround receptive field structure that implements it — not to any actual additional change in physical luminance at the step edge. Retinal ganglion cells and neurons in the lateral geniculate nucleus don't respond only to the light falling on the small patch of retina at the exact center of their receptive field; they're also modulated, in the opposite direction, by light falling on a surrounding annular region — and that's precisely where edge enhancement originates.

Why it happens

A typical visual neuron's receptive field has a concentric structure: a center region and a surrounding ring respond to light in opposite directions — light on the center excites the neuron while light on the surround inhibits it (or vice versa). When both center and surround are illuminated uniformly — say the whole receptive field sits inside a band of uniform luminance — the surround's inhibition roughly cancels the center's excitation, and the neuron's net output settles at a stable, moderate level. That's why the interior of a band looks flat, with no extra exaggeration of brightness or darkness. But when a receptive field happens to straddle a luminance step edge, the symmetry breaks: if the neuron's center sits on the bright side while part of its surround falls on the dark side, the surround delivers less inhibition than it would if the whole receptive field sat inside the bright band — the excitation isn't fully cancelled, so the neuron's output is actually higher than it would be for a receptive field entirely within the uniform bright band. That's exactly why perceived brightness right next to the bright side of an edge gets pushed even higher. Symmetrically, when the center sits on the dark side while part of the surround falls on the bright side, the surround delivers more inhibition than it would if fully within the dark region, and the neuron's output is pushed even lower — hence the darkening right next to the dark side of an edge. None of this requires any extra assumption: it's the natural behavior of the very same population of neurons that encode luminance contrast and settle to a stable response in uniform regions, when they encounter the asymmetric input of a luminance step. Edge enhancement, and the reduced sensitivity to large uniform patches, are two outcomes of the same center-surround antagonistic structure under different luminance distributions.

Studying it

• Receptive-field mapping: recording from single retinal ganglion cells or LGN neurons with microelectrodes while probing them with spots of light at different positions and sizes, plotting the spatial extent of the excitatory and inhibitory zones point by point to directly verify the center-surround structure and its shape.
• Neural recording under step stimuli: positioning an artificial luminance step edge at different locations relative to a receptive field and recording how firing rate changes with edge position, then comparing this against psychophysically measured Mach-band strength to check whether the neural-level overshoot matches the subjective illusion's magnitude.
• Typical independent variables: the position of the spot or edge relative to the receptive field center, the relative size of the spot and the surrounding ring, stimulus contrast.
• Typical dependent variables: neuronal firing rate, the relative weighting of center versus surround (antagonism strength).

Where it stops holding

• This center-surround structure has been directly measured mainly in early visual pathways such as the retina and LGN; receptive fields in higher visual cortex are more complex, so the same center-surround model can't simply be assumed to fully explain edge enhancement at every processing level.
• Antagonism strength (the surround's weight relative to the center) varies between individuals and across retinal locations — receptive field size and center-surround ratio differ between the fovea and periphery, so the same luminance step produces uneven edge enhancement depending on where it falls on the retina.
• This entry describes a static, steadily presented luminance edge. When a stimulus moves quickly or flickers, neurons' temporal response properties come into play as well, making edge enhancement more complex — that falls outside the static case discussed here.

Related

• Same group: A1.31.1 The visual system actively boosts contrast at luminance step edges, producing Mach bands · A1.31.3 Mach banding in smooth gradients gets misread as color banding or posterization defects · A1.31.4 Adding smoother transitions or dither noise weakens edge-enhancement artifacts
• Nearby: A1.04 Contrast Sensitivity and Spatial Frequency Channels
• Search terms: lateral inhibition · center-surround receptive field · on-center off-surround · retinal ganglion cell