Super-Resolution of Thermal Images Using an Automatic Total Variation Based Method

Abstract

The relatively poor spatial resolution of thermal images is a limitation for many thermal remote sensing applications. A possible solution to mitigate this problem is super-resolution, which should preserve the radiometric content of the original data and should be applied to both the cases where a single image or multiple images of the target surface are available. In this perspective, we propose a new super-resolution algorithm, which can handle either single or multiple images. It is based on a total variation regularization approach and implements a fully automated choice of all the parameters, without any training dataset nor a priori information. Through simulations, the accuracy of the generated super-resolution images was assessed, in terms of both global statistical indicators and analysis of temperature errors at hot and cold spots. The algorithm was tested and applied to aerial and terrestrial thermal images. Results and comparisons with state-of-the-art methods confirmed an excellent compromise between the quality of the high-resolution images obtained and the required computational time.

Multi-Scale Ensemble Learning for Thermal Image Enhancement

“…We compared the representative network with the three-level sub-networks to the bicubic method and some other state-of-the-art methods: TEN [10], TIECNN [11], ASR2_s [15], and TIR-DCSCN [12]. The evaluation was done on the test dataset, while the PSNR and SSIM against the ground truth were used as the evaluation metrics.…”

“…The evaluation was done on the test dataset, while the PSNR and SSIM against the ground truth were used as the evaluation metrics. In addition, line profiling [15] was conducted for the in-depth analysis of the pixel values on a reference line.…”

“…Compared to the noisy or blurry edges obtained by the compared methods, our method recovers as much detail as the ground truth image has (see Figure 13). Lastly, to analyze the pixel values of the enhanced thermal images, we performed line profiling [15] on the test images with a scale factor of four. This metric was used to directly analyze and compare the thermal information to the ground truth.…”

“…Several approaches have been proposed to improve a low-quality thermal image by only leveraging software-based techniques. The approaches include detail enhancement (super-resolution) [10][11][12][13][14][15], noise reduction [16][17][18][19], contrast enhancement [20][21][22][23], etc.…”

“…Furthermore, the authors released thermal datasets captured with three different devices for low-resolution, mid-resolution, and high-resolution thermal images. Cascarano et al [15] presented a super-resolution algorithm for aerial and terrestrial thermal images, which was based on total variation regularization. This method is a fully automatic method with a low-cost adaptive rule.…”