Photometric Depth Super-Resolution

Haefner, Bjoern; Peng, Songyou; Verma, Alok; Quéau, Yvain; Cremers, Daniel

doi:10.1109/tpami.2019.2923621

Cited by 28 publications

(15 citation statements)

References 102 publications

(252 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Beginning with the work of B. Horn [1,2,3], Shape from Shading (SfS), algorithms enjoy a long history in computing science research [4,5,6,7,8], continue to attract theoretical investigation in the field of computer vision [9,10,11,12,13,14,15,16,17,18,19,20] and find many practical applications, e.g. in planetary research (Photoclinometry), industrial automation, medical imaging or face recognition [21,22,23,24,25,26,27,28,29].…”

Section: Introductionmentioning

confidence: 99%

Fast Quadratic Shape From Shading With Unknown Light Direction

Yalpani¹

2021

Preprint

View full text Add to dashboard Cite

An algorithm is proposed that extracts 3D shape from shading information in a digital image. The algorithm assumes that there is only a single source of light producing the image, that the surface of the shape giving rise to the image is Lambertian (matte) and that its shape can be locally approximated by a quadratic function. Previous work shows that under these assumptions, robust shape from shading is possible, though slow for large images because a non-linear optimization method is applied in order to estimate local quadratic surface patches from image intensities. The work presented here shows that local quadratic surface patch estimates can be computed, without prior knowledge of the light source direction, via a linear least squares optimization, thus greatly improving the algebraic complexity and run-time of this existing algorithms.

show abstract

Section: Introductionmentioning

confidence: 99%

Fast Quadratic Shape From Shading With Unknown Light Direction

Yalpani¹

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Existing research reveals two major classes of stateof-the-art approaches to depth super-resolution (SR): data-driven methods based on deep neural networks (Riegler et al, 2016;Hui et al, 2016;Voynov et al, 2019) and variational optimization-based approaches (Haefner et al, 2018;Haefner et al, 2020). Among these, learning-based methods bring the promise of leveraging powerful data-driven priors by learning these directly from data, which has proven to achieve impressive quantitative performance; however, as deep networks are trained by optimizing their target functions in an averaged sense, they are likely to produce imperfect estimates for specific (unseen) test instances.…”

Section: Introductionmentioning

confidence: 99%

Can We Use Neural Regularization to Solve Depth Super-Resolution?

Milena¹,

Voynov²,

Artemov³

et al. 2021

Preprint

View full text Add to dashboard Cite

Depth maps captured with commodity sensors often require super-resolution to be used in applications. In this work we study a super-resolution approach based on a variational problem statement with Tikhonov regularization where the regularizer is parametrized with a deep neural network. This approach was previously applied successfully in photoacoustic tomography. We experimentally show that its application to depth map super-resolution is difficult, and provide suggestions about the reasons for that.

show abstract

“…[ 17 , 18 ]. In recent years, the deep-learning (DL) technique has become a research hotspot in various fields, such as object classification and segmentation [ 19 , 20 ], super-resolution [ 21 , 22 ], image denoising [ 23 , 24 ], medical image reconstruction [ 25 , 26 ], etc. In addition to the above applications, it is also adopted in radar signal-processing applications.…”

Section: Introductionmentioning

confidence: 99%

Millimeter-Wave Array Radar-Based Human Gait Recognition Using Multi-Channel Three-Dimensional Convolutional Neural Network

Jiang

Zhang

Yang

et al. 2020

Sensors

View full text Add to dashboard Cite

At present, there are two obvious problems in radar-based gait recognition. First, the traditional radar frequency band is difficult to meet the requirements of fine identification with due to its low carrier frequency and limited micro-Doppler resolution. Another significant problem is that radar signal processing is relatively complex, and the existing signal processing algorithms are poor in real-time usability, robustness and universality. This paper focuses on the two basic problems of human gait detection with radar and proposes a human gait classification and recognition method based on millimeter-wave array radar. Based on deep-learning technology, a multi-channel three-dimensional convolution neural network is proposed on the basis of improving the residual network, which completes the classification and recognition of human gait through the hierarchical extraction and fusion of multi-dimensional features. Taking the three-dimensional coordinates, motion speed and intensity of strong scattering points in the process of target motion as network inputs, multi-channel convolution is used to extract motion features, and the classification and recognition of typical daily actions are completed. The experimental results show that we have more than 92.5% recognition accuracy for common gait categories such as jogging and normal walking.

show abstract

Photometric Depth Super-Resolution

Cited by 28 publications

References 102 publications

Fast Quadratic Shape From Shading With Unknown Light Direction

Fast Quadratic Shape From Shading With Unknown Light Direction

Can We Use Neural Regularization to Solve Depth Super-Resolution?

Millimeter-Wave Array Radar-Based Human Gait Recognition Using Multi-Channel Three-Dimensional Convolutional Neural Network

Contact Info

Product

Resources

About