What is SFR? Spatial Frequency Response in Camera Testing

A camera does not have a sharpness. It has a curve.

That single idea is the thing most people are missing when they ask “how sharp is this camera?” expecting one number back. The honest answer is that a camera reproduces coarse detail almost perfectly and fine detail progressively worse, and the rate at which it gives up on fine detail is its sharpness. SFR — spatial frequency response — is the curve that captures exactly that, and it’s the foundation under nearly every resolution and sharpness metric in imaging.

This is the plain-language explainer we wish existed when we started. We’ll build it from intuition, not from Fourier transforms — though the transform shows up, and we’ll explain it when it does. By the end you’ll be able to read an SFR curve, know the difference between SFR and MTF, understand how it’s measured, and answer the question everyone actually wants answered: what counts as a good value?

This piece sits underneath our Complete Guide to ISO 12233 Test Charts — SFR is the concept, the ISO 12233 chart is the tool you measure it with.

What is SFR?

SFR (Spatial Frequency Response) is a measurement of how well an imaging system preserves contrast as detail gets finer. It’s expressed as a curve: the response starts near 100% for coarse detail and falls toward 0% as spatial frequency increases. The faster the curve falls, the softer the image. SFR is the standard, repeatable way to quantify the sharpness of a camera, lens, or complete imaging system.

That’s the whole concept in one paragraph. Everything below is the why and the how.

First, what is “spatial frequency”?

Imagine a picket fence. Standing close, you see distinct slats with clear gaps — high contrast between white slat and dark gap. Walk backwards, and at some distance the slats blur together into a flat gray. Nothing about the fence changed; what changed is how fine the pattern is in your field of view.

Spatial frequency is a measure of that fineness — how many light-dark cycles fit into a given distance. Low spatial frequency = coarse, widely-spaced detail (a few big slats). High spatial frequency = fine, tightly-packed detail (many thin slats). The “frequency” language is borrowed from sound and signal processing: just as audio has low and high pitches, an image has coarse and fine spatial patterns.

Every real image is a mixture of spatial frequencies stacked on top of each other — the broad shapes are low frequency, the textures and edges are high frequency. A camera that handles all frequencies equally would be perfectly sharp. No camera does. SFR is the map of which frequencies it handles well and which it loses.

How to read an SFR curve

An SFR curve plots response (0 to 1, or 0% to 100%) on the vertical axis against spatial frequency on the horizontal axis. It always starts at or near 1.0 at zero frequency (a uniform gray field has perfect “contrast preservation” — there’s nothing to lose) and falls as frequency rises.

A few reference points let you read any SFR curve at a glance:

• The height of the curve at any frequency tells you how much of the original contrast survives at that level of detail. 0.5 means half the contrast made it through; 0.1 means only a tenth did.
• MTF50 / SFR50 — the frequency at which the curve drops to 0.5 (50% response). This is the everyday “sharpness” number. Higher MTF50 = sharper. It correlates well with how sharp an image looks to people.
• MTF10 / SFR10 — the frequency at which response drops to 0.1 (10%). Often treated as the practical limiting resolution — the finest detail the system can still meaningfully reproduce.
• The shape of the fall-off. A curve that stays high then drops steeply behaves differently from one that sags early but trails off gently — even if they share the same MTF50. Two cameras with identical MTF50 can render very differently.

A practical reading: if someone says “this lens hits MTF50 at 2000 LW/PH,” they mean the lens still preserves half its contrast at a detail fineness of 2000 line widths per picture height — and you can compare that directly against another lens measured the same way.

SFR vs MTF: are they the same thing?

This trips up almost everyone, so here’s the clean version.

SFR (Spatial Frequency Response) is the general, parent term: any measurement of response versus spatial frequency. MTF (Modulation Transfer Function) is a specific kind of SFR — the one measured with true sine-wave (harmonic) inputs, describing modulation transfer in the strict optical sense.

The relationship: every MTF is an SFR, but not every SFR is, strictly, an MTF. An edge-based measurement is “only” an SFR; a sine-based measurement can correctly be called MTF. In everyday engineering the two words get used interchangeably, and modern practice (following ISO 12233) is to say “SFR” as the default term unless you specifically mean the sine-wave variety.

So when you see MTF50 and SFR50 used to mean the same thing, that’s not an error — in the edge-based world most people work in, they’re the same measurement. Just know that the precise distinction exists, because it matters when you compare edge-based and sine-based results (more on that below).

How is SFR actually measured? The slanted-edge method

The dominant technique — the one ISO 12233 standardised and the one almost every tool uses — is the slanted-edge method. It’s clever, and worth understanding because half of “bad SFR” results come from misunderstanding it.

Here’s the chain, in plain terms:

1. You photograph a sharp, straight edge (black-to-gray transition) tilted about 5° from vertical. The tilt is essential — we’ll see why in a second.
2. The software reads across the edge to recover the edge spread function (ESF) — how the perfectly sharp edge got smeared into a gradual transition by the camera. A perfect camera would show an instant step; a real one shows a gentle ramp. The width of that ramp is the blur.
3. It differentiates the ESF to get the line spread function (LSF) — mathematically, the response to an infinitely thin line. (The derivative of a step is a spike; the derivative of a smeared step is a smeared spike.)
4. It runs a Fourier transform on the LSF. The Fourier transform converts the blur shape into a frequency spectrum — and that spectrum is the SFR curve. This is the one piece of math at the heart of it: the shape of the blur and the SFR curve are two views of the same information.

Why the 5° tilt? A camera’s pixels are a coarse grid. If the edge were perfectly vertical, every pixel row would sample it at the same place, and you’d be stuck at pixel-level resolution. Tilting the edge slightly means each row of pixels crosses the edge at a slightly different sub-pixel position. Combine many rows and you reconstruct the edge profile at far finer resolution than a single pixel — “oversampling.” The 5° tilt is what buys you sub-pixel accuracy. Perfectly vertical or horizontal edges break the method.

This is also why the setup details matter so much: if the edge isn’t sharp to begin with (a printed chart with ink bleed), or the chart isn’t flat, or focus is off, the smear the algorithm measures isn’t the camera’s — it’s the chart’s or the setup’s, and it lands in your SFR curve as if the camera did it.

e-SFR vs s-SFR: edge-based vs sine-based

There are two ways to get an SFR, and modern testing uses both because they catch different things.

e-SFR (edge-based SFR) is the slanted-edge method above. It’s fast, robust, and works on a tiny region of the image, so you can measure SFR at dozens of points across the field from a single capture. It’s the workhorse.

s-SFR (sine-based SFR) uses sine-wave patterns — in ISO 12233, a Siemens-star pattern. It’s closer to the textbook definition of MTF and is the right tool for measuring texture loss and revealing the behaviour of scene-adaptive image processing.

Why you’d want both: modern cameras (especially smartphones) run aggressive computational pipelines that sharpen edges while smearing texture. An edge-based measurement sees the sharpened edge and reports a flattering SFR. The sine star sees the smeared texture and reports the truth. Measuring only e-SFR on a heavily-processed camera can make it look sharper than it renders real scenes. The two methods together triangulate what’s actually happening.

Related:Standard vs Enhanced ISO 12233 Chart — the sine star is one of the features only the Enhanced chart carries.

The one conversion worth memorising: 1 line pair = 2 line widths, so a feature at 2000 LW/PH equals 1000 lp/PH. LW/PH is tied to picture height rather than chart height, which is what lets you compare a 12 MP and a 100 MP camera meaningfully — the unit normalises away the framing.

Nyquist, aliasing, and why SFR above Nyquist is suspicious

The sensor’s pixel grid sets a hard ceiling on the detail it can honestly capture: the Nyquist frequency, equal to 0.5 cycles/pixel. Detail finer than Nyquist can’t be reproduced faithfully — instead it aliases, masquerading as false lower-frequency patterns (moiré, those rainbow shimmer artifacts on fine fabrics).

Two practical consequences for reading SFR:

• Legitimate response should be falling toward zero as you approach Nyquist. A well-behaved system has low SFR at Nyquist.
• High SFR beyond Nyquist is a red flag, not a bragging point. It usually means aliasing (false detail) or in-camera sharpening (artificial contrast) — not real resolution. And an SFR value above 1.0 (100%) anywhere is physically impossible for a real optical signal; it’s a dead giveaway that sharpening is active. (We cover that failure mode in the Common Mistakes guide.)

What’s a “good” SFR value?

The question everyone asks, and the honest answer is it depends on what you’re comparing — but here are the anchors that make the number meaningful:

• A good measurement reaches a low reprojection of error and a clean curve, with a well-made chart and proper setup. If your curve is noisy or ratty above mid-frequency, suspect the setup before the camera.
• MTF50 is your sharpness comparison number. There’s no universal “pass” value — it’s relative to the sensor’s pixel pitch and your application. What matters is comparing like-for-like: same chart, same setup, same units.
• Watch the shape, not just MTF50. A camera with high MTF50 but a curve propped up by in-camera sharpening can look worse in real photos than a camera with lower MTF50 and an honest curve. Sharpening inflates the number without adding real detail.
• For relative work, consistency beats absolute targets. If you’re tracking whether a production camera has drifted, you care that today’s curve matches your golden-sample curve — the absolute MTF50 matters less than the stability of the comparison.

The most useful framing: SFR is a comparison tool. A single SFR number in isolation tells you little; an SFR curve measured consistently against a reference tells you almost everything.

Why SFR depends on more than the lens

A final, important point: SFR characterises the whole system, not just the lens. The same lens will give you different SFR curves depending on:

• The sensor — pixel pitch, microlenses, anti-aliasing filter.
• The image processing — sharpening, noise reduction, and tone curves all reshape the curve, sometimes dramatically.
• Focus and aperture — defocus drops the whole curve; diffraction at small apertures softens it too.
• The measurement setup — chart flatness, lighting, and chart quality. A bad chart or setup adds its own blur, which the algorithm cannot separate from the camera’s.

This is why a controlled, repeatable measurement environment matters as much as the camera under test. The point of a standardised chart and procedure is to hold everything except the thing you’re measuring constant — so the SFR curve reflects the camera, not your test bench.

_{The tool for this: a properly made ISO 12233 chart is what gives the slanted-edge algorithm a clean, sharp, flat edge to work from. See why a printed chart undermines this.}

Frequently asked questions

What does SFR stand for? SFR stands for Spatial Frequency Response. It’s a measurement of how well an imaging system preserves contrast as image detail gets finer, expressed as a curve that starts near 100% for coarse detail and falls toward 0% as spatial frequency increases. It is the standard way to quantify camera and lens sharpness.

What is the difference between SFR and MTF? SFR is the general term for response versus spatial frequency; MTF (Modulation Transfer Function) is the specific case measured with sine-wave inputs. Every MTF is an SFR, but an edge-based measurement is “only” an SFR. In everyday use the terms are interchangeable, and ISO 12233 uses “SFR” as the default parent term.

What is MTF50? MTF50 is the spatial frequency at which an imaging system’s response drops to 50% (0.5). It’s the most common single-number sharpness metric because it correlates well with how sharp an image looks to people. A higher MTF50 means the system preserves contrast to finer detail — i.e., a sharper image.

How is SFR measured? Most commonly with the slanted-edge method: photograph a sharp edge tilted ~5°, measure how the camera smeared it (the edge spread function), differentiate to get the line spread function, then apply a Fourier transform to produce the SFR curve. The 5° tilt enables sub-pixel oversampling, giving resolution finer than the pixel grid.

What is the Nyquist frequency in camera testing? The Nyquist frequency is 0.5 cycles/pixel — the hard limit on detail a sensor can honestly capture. Detail finer than Nyquist can’t be reproduced and instead aliases into false patterns (moiré). Legitimate SFR should be falling toward zero near Nyquist; high response beyond it usually indicates aliasing or in-camera sharpening, not real resolution.

Can SFR be greater than 100%? No — not for a real optical signal. An SFR value above 1.0 (100%) is physically impossible without artificial enhancement. If you measure it, in-camera sharpening or edge enhancement is still active and inflating the result. Re-measure with all processing set to neutral and capture in RAW for a valid curve.

Why do I need a special chart to measure SFR? The slanted-edge algorithm assumes the edge it’s analysing was perfectly sharp and flat to begin with, so any blur it measures is attributable to the camera. A precision chart guarantees a sharp, high-accuracy, flat edge; a printed or warped chart adds its own blur that the algorithm can’t separate from the camera’s, corrupting the SFR curve.

Measure SFR you can trust.

SFR is only as good as the edge you measure it from. Calibvision ISO 12233 resolution & SFR test charts are laser-drawn on transmissive film (±15 µm) or photographic paper (±0.1 mm), in 17 sizes and in Standard and Enhanced (Pro) versions — giving the slanted-edge algorithm a sharp, flat, accurate edge to work from. Each ships with a serial-numbered inspection report; CNAS-accredited calibration available on request.

→ See the ISO 12233 SFR test chart range

New to this topic? Start with the Complete Guide to ISO 12233 Test Charts. Ready to measure? Make sure your setup isn’t the bottleneck — read 6 Common Mistakes That Ruin Your ISO 12233 Test Results.