Thermal Image Super-Resolution Challenge - PBVS 2021

Abstract

This paper summarizes the top contributions to the first challenge on thermal image super-resolution (TISR), which was organized as part of the Perception Beyond the Visible Spectrum (PBVS) 2020 workshop. In this challenge, a novel thermal image dataset is considered together with stateof-the-art approaches evaluated under a common framework. The dataset used in the challenge consists of 1021 thermal images, obtained from three distinct thermal cameras at different resolutions (low-resolution, mid-resolution, and high-resolution), resulting in a total of 3063 thermal images. From each resolution, 951 images are used for training and 50 for testing while the 20 remaining images are used for two proposed evaluations. The first evaluation consists of downsampling the low-resolution, midresolution, and high-resolution thermal images by ×2, ×3 and ×4 respectively, and comparing their super-resolution results with the corresponding ground truth images. The second evaluation is comprised of obtaining the ×2 superresolution from a given mid-resolution thermal image and comparing it with the corresponding semi-registered highresolution thermal image. Out of 51 registered participants, 6 teams reached the final validation phase.

A Novel Domain Transfer-Based Approach for Unsupervised Thermal Image Super-Resolution

“…The cameras were mounted on a panel, trying to minimize the baseline distance between the optical axis to obtain an almost registered image set. This dataset was used as a benchmark in the first and second thermal image super-resolution challenge organized on the workshop Perception Beyond the Visible Spectrum of CVPR2020 [ 17 ] and CVPR2021 conferences [ 18 ].…”

“…At the second thermal image super-resolution challenge [ 18 ], different teams also participated and presented their approaches. For this second challenge, the same dataset is used, but for Evaluation 1, just on HR were considered and Evaluation 2 maintains the same method (MR to HR images).…”

“…As proposed in [ 16 ], and adopted in the PBVS-CVPR2021 challenge [ 18 ] (referred to as a Evaluation 2 ), the quantitative evaluation of the approach presented in the current work is performed by means of the average PSNR and SSIM measures between the generated SR image and the semiregistered HR counterpart obtained from the other camera; this evaluation is illustrated in Figure 6 . Due to the camera baseline, the information in the images is not the same; hence, just an ROI of the 80% of the image size, centered at each image, is considered.…”

“…Due to this lack of thermal images, a novel dataset was proposed in [ 16 ] with three different resolutions (low, mid, and high) obtained with three different thermal cameras. This dataset has been used as a benchmark in the first and second thermal image super-resolution challenge on PBVS-CVPR2020 [ 17 ], and PBVS-CVPR2021 ( ) (accessed on 12 December 2021) [ 18 ] workshops, where several teams around the world have participated, and a baseline has been obtained.…”

XCycles Backprojection Acoustic Super-Resolution

“…To the best of the authors' knowledge, this is the first work to provide such a large dataset with its specific characteristics for the SISR problem and the sub-sampling phase delay error problem; • The proposed benchmark and the developed method outperformed the classical interpolation operators and the recent feedforward state-of-the-art models and drastically reduced the sub-sampling phase delay error estimation. • The proposed model contributed to the Thermal Image Super-Resolution Challenge-PBVS 2021 [11] and won the first place with superior performance in the second evaluation when the LR image and HR image are captured with different cameras.…”