An Evaluation of Objective Image Quality Assessment for Thermal Infrared Video Tone Mapping

Teutsch, Michael; Sedelmaier, Simone; Moosbauer, Sebastian; Eilertsen, Gabriel; Walter, Thomas

doi:10.1109/cvprw50498.2020.00062

Cited by 6 publications

(5 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…While general tone mapping for TIR images is still a niche topic 5 , deep learning-based tone mapping for TIR images is almost non existent in the literature. Teutsch et al 1 successfully trained the Context Aggregation Network (CAN) architecture, based on the works of Chen et al 8 , to approximate the Realtime TMO 16 in a supervised manner.…”

Section: Related Workmentioning

confidence: 99%

“…This allows for a finer discretization of those values and therefore a higher level of detail. This is especially important in Thermal Infrared (TIR) imaging, where only one intensity channel is available and therefore a high number of bpp is needed to maximize the contrast 1 . Showing this content on standard displays, however, comes with its own challenges as most monitors are limited to only 8 bpp.…”

Section: Introductionmentioning

confidence: 99%

“…In this paper, such TMOs are investigated for the compression of TIR images and videos with 16 bpp. There are actually very few methods available at this time specifically dealing with tone mapping in the TIR spectrum 1,5 . Inspired by Luo et al 5 , the Multi-Scale Retinex (MSR) algorithm, originally intended for contrast enhancement 6 , is implemented and optimized as a reference for this purpose (see Fig.…”

Section: Introductionmentioning

confidence: 99%

“…Extensive experiments on the public FLIR ADAS Dataset * provide a comprehensive comparison with the state-of-the-art and prove the effectiveness of our proposed method. Our contributions can be summarized as (1) an optimization of the MSR algorithm to generate a reference TMO for TIR images and videos, (2) an approximation of this reference TMO with a deep learning-based U-Net neural network architecture in a supervised training scheme, (3) the injection of self-supervised denoising for joint multi-task learning of tone mapping and denoising within the U-Net architecture, and (4) a deep discussion on self-supervised denoising and deflickering. Each part is analyzed in detail and accompanied by comprehensive experiments and results on the public FLIR ADAS Dataset 12 .…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Joint tone mapping and denoising of thermal infrared images via multi-scale Retinex and multi-task learning

Goedrich,

Koenig,

Eilertsen

et al. 2023

Infrared Technology and Applications XLIX

View full text Add to dashboard Cite

Cameras digitize real-world scenes as pixel intensity values with a limited value range given by the available bits per pixel (bpp). High Dynamic Range (HDR) cameras capture those luminance values in higher resolution through an increase in the number of bpp. Most displays, however, are limited to 8 bpp. Naïve HDR compression methods lead to a loss of the rich information contained in those HDR images. In this paper, tone mapping algorithms for thermal infrared images with 16 bpp are investigated that can preserve this information. An optimized multi-scale Retinex algorithm sets the baseline. This algorithm is then approximated with a deep learning approach based on the popular U-Net architecture. The remaining noise in the images after tone mapping is reduced implicitly by utilizing a self-supervised deep learning approach that can be jointly trained with the tone mapping approach in a multi-task learning scheme. Further discussions are provided on denoising and deflickering for thermal infrared video enhancement in the context of tone mapping. Extensive experiments on the public FLIR ADAS Dataset prove the effectiveness of our proposed method in comparison with the state-of-the-art.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Joint tone mapping and denoising of thermal infrared images via multi-scale Retinex and multi-task learning

Goedrich,

Koenig,

Eilertsen

et al. 2023

Infrared Technology and Applications XLIX

View full text Add to dashboard Cite

show abstract

“…B IT depth expansion (BDE) is an emerging issue to generate and display a high bit-depth (HBD) image from the low bit-depth (LBD) source. With the rapid development of television technology, there are numerous ultra high-definition display terminals supporting 10-bit or even 12-bit luminance levels, while the mainstream video and image sources are just 8-bit [34]. Thus, expanding the bit depth becomes a popular topic for high-dynamic-range (HDR) display [22], display acceleration on mobile (SoC) [8] and video compression [9], [24], [43].…”

Section: Introductionmentioning

confidence: 99%