LAUSR

Estimation of vital signs using image processing techniques have already been proved tohave a potential for supporting remote medical diagnostics and replacing traditional measurementsthat usually require special hardware and electrodes placed on a body. In this paper, we furtherextend studies on contactless Respiratory Rate (RR) estimation from extremely low resolution thermalimagery by enhancing acquired sequences using Deep Neural Networks (DNN). To perform extensivebenchmark evaluation, we acquired two thermal datasets using FLIR® cameras with a spatialresolution of 80 × 60 and 320 × 240 from 71 volunteers in total. In-depth analysis of the proposedConvolutional-based Super Resolution model showed that for images downscaled with a factor of 2and then super-resolved using Deep Learning (DL) can lead to better RR estimation accuracy thanfrom original high-resolution sequences. In addition, if an estimator based on a dominating peakin the frequency domain is used, SR can outperform original data for a down-scale factor of 4 andimages as small as 20 × 15 pixels. Our study also showed that RR estimation accuracy is betterfor super-resolved data than for images with color changes magnified using algorithms previouslyapplied in the literature for enhancing vital signs patterns.