Thermal imager selection is a balance between 49,152 pixels at 256×192 and a lower NETD such as 40 mK versus 50 mK, rather than a choice between resolution and sensitivity alone.
The decision depends on target angular size, temperature contrast, standoff distance, optics, and whether the camera must produce radiometric measurements or only qualitative images.
Resolution, NETD and the actual detection task
Detector resolution controls how finely a scene is spatially sampled, while NETD expresses the smallest temperature difference the system can resolve under defined test conditions. A 320×240 detector contains 76,800 measurement pixels, and a 640×480 detector contains 307,200, so the latter provides more than 28 times as many spatial samples as a 120×90 detector with 10,800 pixels [S4].
NETD governs visibility of weak thermal contrast, while resolution governs whether two nearby features can be separated. A lower NETD can reveal a faint hot spot, but it cannot recover spatial detail that the detector and optics cannot resolve. Conversely, extra pixels do not guarantee a usable thermal signal when scene contrast is below the noise level.
For target detection, compare the target's angular dimensions with the detector's spatial sampling. For temperature measurement, do not treat one output pixel as a valid spot reading. A measurement spot should cover at least a 3×3 pixel cluster, which is the stated basis for a measurement field of view rather than an instantaneous single-pixel reading [S4].
Compare the main selection options
A 160×120 camera with 40 mK NETD gives 19,200 detector pixels, while a 256×192 camera with 50 mK NETD gives 49,152 pixels. The 256×192 option samples a target with 2.56 times as many pixels, while the 160×120 option detects temperature differences 10 mK more sensitively [S2].
The comparison is conditional: the 160×120 versus 256×192 example is a user selection case rather than a universal ranking [S2], while 320×240 and 640×480 are presented as practical inspection points [S3][S4].
Where sensitivity takes priority

Choose lower NETD when the operating scene contains weak, nearly equal-temperature targets, such as heat leakage, early-stage electrical faults, or a warm object against a similarly warm background. The 40 mK versus 50 mK difference is modest, so it should not outweigh geometry unless the task is deliberately near the detection threshold [S2].
Low NETD is particularly important in search and rescue, where a faint heat signal must be detected across a wide field of view [S3]. For broad surveillance, sensitivity may be more valuable than fine spatial detail, but a wide lens reduces the target's angular size and can therefore offset some of that advantage.
When selecting a thermal camera, consider thermal sensitivity (NETD), lens choice, durability, and the detail and distance required for the intended application [S3].
Where resolution takes priority
Choose higher resolution when the target is small, the camera is far away, or the diagnostic question requires the outline of an anomaly rather than merely its presence. For electrical panels and large equipment, 320×240 is described as a practical starting point, while 640×480 provides more detail for advanced troubleshooting [S3].
Increasing pixel count helps only if each pixel receives usable signal. Reducing pixel dimensions to improve spatial sharpness also reduces the signal captured per pixel and can lower the signal-to-noise ratio, making smaller-pixel detector development technically difficult and costly [S4]. This is the physical reason sensitivity and resolution compete within a constrained optical and detector budget.
Standoff also changes the result. Increasing distance makes the target subtend fewer detector pixels and can alter measured temperatures, so physical resolution or optical tools may be needed to preserve measurement precision [S4]. Detailed treatment of small hot spots is covered in Thermal Sensor Resolution for Small Hot Spot Detection.
Verify the resolution and optics claims

Require the native detector array, not the exported image size.
The thermal imager category page is a useful reference when separating detector specifications from complete-camera performance.
Accuracy, radiometry and measurement limits
NETD is not a substitute for temperature accuracy. NETD describes noise-limited sensitivity, while accuracy requires a defined calibration, a traceable temperature reference, an appropriate emissivity assumption, and control of reflections and atmospheric effects. A more sensitive detector can still provide an inaccurate absolute temperature if the measurement chain is not calibrated. [S4]
Radiometric capability adds constraints because accurate temperature estimates are not simply a by-product of visual sensitivity [S2]. For electrical inspection, request spot and area measurements, emissivity adjustment, reflected-temperature correction where applicable, and a defined target-size rule. Use the camera's full measurement field of view, not a single pixel, when reporting a temperature [S4].
NFPA 70B and NFPA 70E cover electrical maintenance and inspection considerations, while ASTM C1060 and ASTM E1186 apply to building-envelope infrared inspection [S4]. These standards do not replace the core optical decision: calculate whether the target occupies enough pixels and whether its expected contrast exceeds the camera's effective noise at the installed lens and distance.
Practical selection by use case

For electrical-panel troubleshooting, specify 320×240 or 640×480 when the work involves small components, busbars, or detailed fault localization. Retain a lower NETD option when the anomalies are expected to be weak, but do not pay for resolution that the target size and viewing distance cannot use [S3][S4].
For outdoor inspection, use the target's angular size, not the nominal source resolution alone. A 640×480 array can improve small-target separation, but a wider field of view reduces spatial detail, so a balanced 320×240 configuration may be more suitable for scanning large equipment [S3].
For building-envelope work, low emissivity surfaces and reflection-related temperature errors can dominate the result. Specify radiometric correction, suitable emissivity handling, and an external reference where needed, then verify that the target spans the required measurement area. For electronics work near a bench, the 160×120 at 40 mK versus 256×192 at 50 mK comparison illustrates the trade-off cleanly: retain 40 mK when thermal contrast is weak, and select 256×192 when component geometry matters more [S2].
Procurement checklist and trackable specifications
Before approving a thermal imager, record native array dimensions, pixel pitch, NETD and its test conditions, lens f-number, spectral band, frame rate, temperature range, accuracy specification, minimum focus distance, and whether output is radiometric. Confirm that the stated pixel array is native, the lens supports the stated NETD, and the target fills the required measurement field of view [S1][S2][S4].
The clearest comparison is still the 256×192 at 50 mK versus 160×120 at 40 mK decision: choose 256×192 for geometry and target separation, and choose 40 mK for weak-contrast detection [S2]. Validate the choice with representative targets at the intended distance, because exported resolution, upscaling, lens aperture, optical focus, and processing can change the result materially.