As far as I know - video superresolution is doing continues multiframe superresolution.
so it looks at the past 4 frames. and when your imager does 9hz - you can easily get ghosting artifacts on moving targets like cars or people.
Thermal Camera+ does not support video capture. It also lacks any timer or intervall settings which could subsititude video. My main concern with the app right now is that it burns in the min and max temperature as well as the palette gradient and to archive that - images will get upscaled to 508x640 - either using the Nearest Neighbors setting - basically making every pixel become 4 pixels - or the bilinear upscale by mixing the additional pixels form either a 120x160 supperresolution picture or the 60x80 native picture.
Supperresolution has two sharpening modes - while "Medium" is giving more accurate readings, as well as looking better for NN and bilinear upscale the "Harder" sharpening mode will make fine details and edges more visible by inducing more contrast to some edges and more noise to the overall image (I would love to compare it to
Flir DDE which they developped for survaliance applications - using same sensor and same target).
The data burn in basically voids data of a few pixels and therefore should have a setting to be turned off or be moved off the image and into the 28 wide column of pixels for the gradient.
I did not succeed to make a imagemagick script to convert a image out of the Thermal Camera+ app into a .jpg without burned in gradient or data to use for photogrammetry or stitching.
I know that it is possible to somehow extract a raw thermal image from FLIR radiometric .jpeg with some sript found on the forum but I failed to do so as well. and the FLIR firmware upsclaes the images from the CAT S60 lepton - or Flir ONE - to 240x320 "IR resolution" in flir tools which is 2
4x (= 16x) the effective resolution. As well as to a .jpg resolution of 480x640(which could be due to the to bad resolution and still downscaled and cropped visula image being overlayed in MSX). Flir also applies histogramm equalization and some kind of streching to apply the whole color range of the palette on only the range of temperature shown in the image. This has the result of images capture with the myFLIR app, of the same target but with different background highs and lows will look different even if I set the range of temperature to the same min and max the whole image will look different because factors like gain and gamma can't be changed(It might be that some of those factors are calculated through the parameters in Flir tools: Emissivity, Refl. temp., Distance, Atmosphereic temp., Ext. optics temp., Ext. optics trans. and Relative humidity).
I am still looking for a solution to capture multiple images - by hand, as interval or as 9hz video - using the very same fake color - or a radiometric fashion where I can adjust the fake color palette in post. A solution would be to turn your all images into a raw array of 80x60 pixels and generate a 14bit greyscale that covers the whole sensor output range in 50mK steps. Flir Tools can generate a 320x240 .cvs with temperature values... but this means they read 16 temperatures out of a single Lepton pixel to 3 decimal places in °C and I know how much of that is bullshit in the end.
Right now Thermal Camera+ with Superresolution on "Harder" (or maybe I should do "Medium" instead) and NN zooming with locked palette(which locks the range of the temperatures as well as gamma) and shooting from multiple angles for photogrammetry or getting closer and shooting an array of a few rows with good overlap for stiching in ICE - this is only effective at a large FoV when the target is flat as perspective becomes an issue with planar motion.
Videos and Timelapses made in myFLIR app are bad because locking the temp range of the is broken due to the Shutter going crazy and unless your burn in a spot you can't get any data. not even a min/max and gradient. Flir radiometric video is only part of their high end research cameras.