Johnson Criteria for Thermal Imaging is a practical method for estimating whether an infrared imaging system has enough spatial detail to detect, recognize, or identify a target at a given range. It is widely used in early EO/IR system design because it connects detector format, pixel pitch, focal length, field of view, target size, and image quality to a mission-level question: what can the user actually distinguish in the image? For OEM engineers, the Johnson Criteria should be treated as a range-screening model, not a guaranteed field result, because real performance also depends on thermal contrast, atmosphere, optics, signal processing, display conditions, motion, and the observer or algorithm interpreting the image.
How Does Johnson Criteria for Thermal Imaging Work?
The Johnson Criteria are based on the number of resolvable line pairs across a target’s critical dimension. A line pair is one bright-dark cycle. In imaging terms, it is a spatial frequency measurement: the more resolvable cycles across the target, the more structure is available for the observer or algorithm to classify the object.
The “critical dimension” is usually the smaller or most task-relevant dimension of the target. For a standing person, height may be used in some calculations, but shoulder width or torso width can be more conservative for identification. For vehicles, the relevant dimension may be width, height, or a specified target aspect depending on viewing angle. This target definition must be fixed before comparing modules or lenses.
The commonly cited Johnson thresholds are approximate task thresholds at about 50% probability under controlled assumptions. Detection requires roughly 1 line pair across the target, recognition roughly 4 line pairs, and identification roughly 6.4 line pairs. With ideal sampling, one line pair requires two pixels, so these values are often converted to about 2 pixels, 8 pixels, and 13 pixels across the target. Some procurement or application-specific models use more conservative values, such as higher pixel counts for recognition or identification. The exact convention matters; a range estimate based on 13 pixels for identification is not directly comparable with a range estimate based on 24 pixels.
The basic geometric relationship is simple. Instantaneous field of view, or IFOV, is approximately pixel pitch divided by focal length. Pixels across the target are approximately target size divided by range and then divided by IFOV. Rearranged, range is approximately target size divided by required pixel count and IFOV. This is why focal length and pixel pitch are often more important to DRI range than detector format alone.
What Parameters Determine Thermal Imaging DRI Range?
Detection, recognition, and identification range are driven by both geometry and contrast. Geometry determines how many pixels or line pairs fall across the target. Contrast determines whether those pixels contain usable information above noise and blur. A thermal camera with enough pixels on target can still fail if the target is nearly the same apparent temperature as the background, if atmospheric absorption reduces contrast, or if the optics and processing suppress the spatial detail needed by the observer.
The first geometric parameters are detector resolution, pixel pitch, focal length, and field of view. A 1280×1024 detector can provide a wider field of view at the same angular sampling, or finer sampling at the same field of view, depending on the lens. A 640×512 detector with a longer focal length can exceed the long-range DRI performance of a higher-resolution detector with a short lens. For compact uncooled LWIR designs, modules such as SPECTRA L06 640×512 LWIR 12μm are often evaluated by matching lens options to the required target size and field of regard.
The second group of parameters is image quality. Modulation transfer function, focus stability, lens transmission, f-number, detector NETD, non-uniformity correction, bad-pixel replacement, sharpening, temporal filtering, and compression can all change the effective detail available at the display or algorithm input. Resolution charts and spatial frequency response methods, such as those addressed by ISO 12233:2024, are useful references for thinking about measured sharpness, even though Johnson Criteria calculations for thermal systems also require IR-specific contrast and sensitivity assumptions.
The third group is radiometric and environmental. Thermal DRI range often assumes a target-to-background temperature difference, a defined atmosphere, humidity, path length, and sometimes a standard target. MWIR and LWIR bands are affected differently by atmospheric conditions and target emission. Cooled MWIR modules, such as SPECTRA M12 1280×1024 Cooled MWIR, are commonly considered where long-range contrast, lower noise, and narrower fields are more important than size, cost, or cooler power.
Johnson Criteria vs DRI: What Is the Difference?
DRI describes the task. Johnson Criteria provide one way to estimate the spatial sampling required for that task. Detection means the user can determine that an object is present. Recognition means the user can classify the object, such as person, vehicle, boat, or animal. Identification means the user can determine a more specific type or confirm target identity to the level required by the application.
This distinction is important because DRI numbers in datasheets can look precise while hiding different assumptions. A stated detection range is incomplete unless the target size, thermal contrast, probability level, atmospheric model, lens, frame rate, display processing, and criterion are known. A “vehicle detection range” based on a large high-contrast target cannot be applied to a partially occluded person or a low-contrast object near the horizon.
Johnson Criteria also should not be confused with complete human perception modeling. The original method is a spatial-detail approximation. It does not fully model target shape, clutter, motion, search time, display size, operator training, image enhancement, or false alarm rate. More advanced target acquisition models may incorporate three-dimensional noise, system MTF, display characteristics, and task probability in more detail. Standards and measurement frameworks such as EMVA 1288 are useful examples of structured camera characterization, but OEM thermal range validation still needs application-specific field data.
For AI-enabled systems, the same caution applies. A neural detector may operate below traditional human recognition thresholds for some targets if it has been trained on similar imagery, but it may also fail under domain shift, weather changes, compression artifacts, or unusual target aspect angles. Systems such as NEXUS LV0619B AI multi-band Ethernet/SDI should therefore be evaluated with both geometric pixels-on-target analysis and dataset-based detection metrics.
When to Use Johnson Criteria for LWIR, MWIR, and Multi-Band Modules
Johnson Criteria are useful in the concept and architecture stages for LWIR, MWIR, SWIR, visible, and fused systems, but the interpretation changes by waveband. In LWIR, uncooled microbolometers are often selected for passive day-night imaging, lower power, simpler integration, and cost-sensitive volume applications. The design trade-off is usually between field of view, lens size, range, and thermal sensitivity. Johnson analysis can quickly show whether a lens-detector combination has enough sampling before deeper thermal modeling is performed.
In cooled MWIR, Johnson Criteria are often used for longer-range surveillance, airborne payloads, and stabilized EO/IR systems. MWIR sensors can offer high sensitivity and strong target contrast in many scenarios, especially with cooled detectors and suitable optics. The trade-off is cooler power, startup time, mechanical envelope, cost, and lifecycle considerations. In these systems, the Johnson calculation is only the first gate; optical MTF, vibration, stabilization residuals, atmospheric transmission, and focus control can dominate real range performance.
In multi-band systems, Johnson analysis should be performed separately for each imaging channel and then interpreted at the fused-system level. Visible or SWIR channels may provide texture, markings, and edges that are not available in thermal imagery, while LWIR or MWIR channels may provide target contrast when visible illumination is poor. A dual-band module such as FUSION LV1225A 1280×1024+2560×1440 requires attention to registration accuracy, latency, field-of-view matching, and how fused images are displayed or processed. A fused image can improve operator understanding, but it does not remove the need for adequate sampling in the channel that carries the target evidence.
How Should OEMs Use Johnson Criteria for Module Selection?
OEMs should use Johnson Criteria as an engineering filter before committing to a detector, lens, and processing architecture. The first step is to define the target set, including physical dimensions, aspect angles, thermal contrast assumptions, required probability, and the exact meaning of detection, recognition, and identification for the product. The second step is to calculate pixels on target across candidate focal lengths and detector pitches. The third step is to apply image-quality and environmental margins rather than treating the geometric range as the delivered product range.
The practical conclusion is that module selection should be based on the complete imaging chain, not on detector resolution alone. A suitable OEM module is the one that meets the required target-acquisition task with margin while also satisfying interface, latency, mechanical, power, radiometric, software, and production constraints. Johnson Criteria help narrow the options; field imagery and application-specific validation confirm the final choice.
FAQ: Johnson Criteria for Thermal Imaging
How many pixels are needed to identify a human with a thermal camera?
A common Johnson-based estimate is about 13 pixels across the target’s critical dimension for identification under idealized assumptions. In real systems, more pixels may be needed because a person can be partially occluded, viewed from an unfavorable angle, blended into clutter, or close in apparent temperature to the background. If identification means confirming equipment, posture, carried objects, or behavior, the required pixel count can be substantially higher.
Why do two thermal cameras with the same resolution have different DRI ranges?
Resolution is only the detector format. DRI range also depends on pixel pitch, focal length, lens quality, optical transmission, focus, detector sensitivity, image processing, target contrast, and atmospheric conditions. A 640×512 camera with a narrow field-of-view lens may place more pixels on a distant target than a 1280×1024 camera with a wide-angle lens.
Can Johnson Criteria predict thermal camera range at night?
Johnson Criteria can estimate the spatial sampling required at night because thermal imaging does not depend on visible illumination. However, the calculation still needs target-to-background thermal contrast and environmental assumptions. Nighttime range may improve or degrade depending on target temperature, background temperature, humidity, rain, fog, and atmospheric path length.
Is Johnson Criteria still useful for AI thermal imaging?
Yes. Johnson Criteria remain useful as a baseline for determining whether the image contains enough spatial information for a task. AI models still need pixels, contrast, and stable image quality. The final AI performance should be measured with representative datasets, including false alarms, missed detections, weather variation, target aspect, range, and sensor processing settings.
