,
just got a Seek Shot Pro ... it seems to sometimes have trouble taking sharp images (especially the stove in the last picture).
Thermal is often like that: sharp edges can be really hard to achieve.
There are so many factors working against one in terms of achieving good picture quality (compared to a mid-range cellphone visible image).
1 - resolution - this is a big issue: you just don't have many pixels, and they are all (necessarily) processed nearly to death in order to get a usable image. A mid-range 320x240 camera has 0.077 megapixels but no self-respecting cellphone boasts anything less than 10 megapixels - that's a factor of
130 times more resolution.
2 - lens quality - whilst it
is possible to create optically perfect lenses at LWIR, it can be horribly expensive. Commodity thermal cameras use the simpler lens designs (eg fewer elements) and are often constructed with cheaper, lower performance materials produced to lower precision. That all helps to keep the device affordable. One of the lenses I have for my FLIR SC-660 lists at about the same price as a small car (and it's not
all down to what I call "FLIR Tax"). Also, thermal camera lenses are normally as large-aperture as possible, within the cost constraints, which means the depth of field is limited. For fixed-focus cameras the manufacturer sets the lens up for roughly the hyperfocal distance and hopes for the best.
3 - wide frequency range - this is a subtle corollary to (2) and, in effect, means you have
no chance of getting sharp images: typical LWIR cameras respond to a wavelength range of 7µm to 14µm, a 2:1 range. It is simply impossible for a simple lens to bring all of those wavelengths into sharp focus at the same time. The best you can hope for is that your subject will emit the majority of its energy in a moderately narrow range, but the nature of blackbody radiation (which we're looking at) means that's unlikely to happen. (In visible light lenses you get 'chromatic aberration', perhaps most often noticed as purple fringing on edges. These days even phone cameras have pretty well-designed multi-element lenses that minimise the effect, but applying the same level of complexity to a thermal camera lens is difficult-to-impossible, not least because there is no LWIR equivalent of the subtly different glass types used in visible lenses).
4 - processing - touched on in (1), but you really do need to understand that thermal imaging sensors are a very different beast from visible light cameras; adjacent pixels on a thermal sensor can have wildly different characteristics. These are mapped individually (pixel by pixel, sensor by sensor) by the manufacturer and for each sensor an individualised 'non-uniformity correction' data set is created - basically a big table of data that describes the offsets, non-linearities, drift characteristics and scaling differences of each individual pixel. Every pixel in every frame from a thermal sensor has its output tweaked according to the corresponding data points in the NUC table. And then there's noise reduction: thermal sensors are noisy, not least because they are responding to infinitesimally tiny variations in temperature. If you have a visible light camera with adjustable ISO, imagine running it at its highest ISO setting
all the time: that's essentially how a thermal camera lives its life. For any given sensor type the only practical way to reduce its
actual noise level is to use a larger-aperture lens. I have germanium lenses as wide as f/0.75; my widest photographic lens is f/1.4 and I believe the widest aperture 35mm camera lens in common production is f/0.95 (Nikon 58mm). Stanley Kubrick famously used one-third of the world's supply of ultra-wide aperture f/0.7 lenses (Carl Zeiss Planar 50mm) during the filming of
Barry Lyndon - but I digress...
Sorry I've warbled on so much (and I've probably missed some important stuff) but hopefully this goes some way to explain some of what's going on. Reading back over what I've written I realise there's a lot more that could be said - for example, I mentioned the hyperfocal distance in (2), whereas it usually isn't appropriate to focus a thermal lens there because the low resolution of the sensor means that anything further than a certain distance will be too small for the sensor to detect, so it makes sense to focus a little nearer than hyperfocal ... and so it goes on.