Contrast Sensitivity: The Spatial Frequencies You Can and Can't See
Your eye is not equally sharp at all scales. It's brilliant at medium detail - a face across a room, text at reading distance - but oddly blind to the very fine and the very coarse. There's a single famous image that reveals this in one glance: stripes that fade in an arch as they get finer, tracing the exact shape of your own sensitivity. This is the interactive guide to the contrast sensitivity function - the map of the detail you can, and can't, see.
Sharpness has a favorite scale
Vision scientists describe detail by spatial frequency: how many light-dark cycles a pattern packs into a degree of your visual field. Fine stripes are high frequency; broad bands are low frequency. And it turns out your eye's ability to detect a faint pattern - its contrast sensitivity - depends strongly on that frequency. You're not a camera with a flat response; you have a favorite scale.
Plot sensitivity against spatial frequency and you don't get a flat line or a simple cliff at the acuity limit. You get an arch: sensitivity rises to a peak around a few cycles per degree, then falls off toward fine detail (the true resolution limit) and toward coarse detail (where the eye's edge-seeking wiring stops caring about slow, smooth changes). The visual system is band-pass - tuned to medium-scale contrast, the scale where the edges and textures that matter usually live.
The chart that draws your vision
Here is the famous demonstration - the Campbell-Robson chart. Stripes get finer from left to right, and their contrast fades from bottom to top. Every point at a given height has the same physical contrast, so if you were a flat detector the stripes would vanish along a straight horizontal line. Instead, the line where they disappear arches up in the middle - you see the faint mid-frequency stripes higher than the fine ones. That arch is your CSF, drawn by your own eye.
The stripes vanish along your sensitivity curve
Frequency increases left→right; contrast decreases bottom→top. Look at where the stripes fade into gray: the boundary is an arch, not a flat line, because you're most sensitive to medium-frequency stripes. Toggle the modelled CSF curve to see it match the arch you perceive. (Sit back a little and the arch shifts.)
Your sensitivity curve
Pull the arch out into a proper graph. Plotting sensitivity against spatial frequency gives the canonical CSF: a peak near 3-5 cycles per degree, a steep fall to the acuity limit around 40-60 cycles per degree (finer than that and no contrast is enough), and a gentler droop at low frequencies. Where you sit on the curve depends on viewing distance - move back and a pattern's frequency rises. Test a single grating against the curve.
Peak in the middle, cliffs on both sides
The contrast sensitivity function as a curve. Choose a grating's spatial frequency and see how much contrast it needs to be visible (1 / sensitivity) - tiny in the sweet spot, huge near the fine-detail cliff. The sample grating on the left is drawn at the threshold contrast for that frequency.
Edges the eye invents
The low-frequency droop has a famous side effect. Because the eye's neurons emphasize differences over absolute levels (lateral inhibition), a staircase of flat gray steps sprouts illusory bright and dark lines at each edge - Mach bands. The steps are perfectly flat; your visual system exaggerates the boundaries. It's the CSF's edge-seeking bias made visible, and a cousin of simultaneous contrast.
Flat steps, imaginary edges
A staircase of uniform gray steps. Look at each boundary: a faint bright line seems to hug the lighter side and a dark line the darker side - Mach bands - though the readout confirms every step is perfectly flat. Add a smooth ramp to compare; the illusory lines are strongest at sharp steps.
Throwing away the invisible
The CSF is money. Image compression like JPEG splits each block into spatial-frequency components and keeps them in proportion to how well you see each one - lavishing bits on the medium frequencies you're sensitive to and brutally discarding the high frequencies you can barely detect. The file shrinks dramatically with little visible change, because it's spending on exactly what your CSF cares about. Cut the high frequencies and judge the damage.
Keep what you see, drop what you don't
A detailed image with a slider that removes its highest spatial frequencies (a low-pass, like aggressive compression). Trim the very finest detail and it's nearly invisible - that's the high-frequency tail of the CSF you can't see. Cut into the medium frequencies and the loss becomes obvious.
The vocabulary
The terms of spatial vision.
Best practices and pitfalls
Test your understanding
Six questions on spatial frequency, the CSF, Mach bands, and compression. Instant feedback, no scores recorded - a wrong answer comes with a short explanation.
Quick check
Continue your journey
Contrast sensitivity connects to chromatic acuity, cone wiring, edge illusions, and compression - here's where to go next.
Chromatic Acuity: Why You See Detail in Brightness, Not Color
The same story for color: even lower spatial resolution.
Vision · 5Human Color Vision: Cones, Opponent Signals, and the Brain
The retinal wiring behind the CSF and lateral inhibition.
Vision · 30Color Illusions and the Limits of Perception
Mach bands join the gallery of visual illusions.
Foundations · 84Simultaneous Contrast: Why a Color Depends on Its Neighbors
The edge-enhancing wiring, working on color.
Digital · 39Color in Film and Video
Where frequency-based compression lives in the pipeline.
Digital · 67Color Quantization and Dithering
Dithering to defeat the banding the CSF exposes.