Depth maps acquired with ToF cameras have a limited accuracy due to the high noise level and to the multi-path interference. Deep networks can be used for refining ToF depth, but their training requires real world acquisitions with ground truth, which is complex and expensive to collect. A possible workaround is to train networks on synthetic data, but the domain shift between the real and synthetic data reduces the performances. In this paper, we propose three approaches to perform unsupervised domain adaptation of a depth denoising network from synthetic to real data. These approaches are respectively acting at the input, at the feature and at the output level of the network. The first approach uses domain translation networks to transform labeled synthetic ToF data into a representation closer to real data, that is then used to train the denoiser. The second approach tries to align the network internal features related to synthetic and real data. The third approach uses an adversarial loss, implemented with a discriminator trained to recognize the ground truth statistic, to train the denoiser on unlabeled real data. Experimental results show that the considered approaches are able to outperform other state-of-the-art techniques and achieve superior denoising performances.

State-of-the-art approaches for the estimation of hyperspectral images (HSI) from RGB data are mostly based on deep learning techniques but due to the lack of training data their performances are limited to uncommon scenarios where a large hyperspectral database is available. In this work we present a family of novel deep learning schemes for hyperspectral data estimation able to work when the hyperspectral information at our disposal is limited. Firstly, we introduce a learning scheme exploiting a physical model based on the backward mapping to the RGB space and total variation regularization that can be trained with a limited amount of HSI images. Then, we propose a novel semi-supervised learning scheme able to work even with just a few pixels labeled with hyperspectral information. Finally, we show that the approach can be extended to a transfer learning scenario. The proposed techniques allow to reach impressive performances while requiring only some HSI images or just a few pixels for the training.

Time-of-Flight data is typically affected by a high level of noise and by artifacts due to Multi-Path Interference (MPI). While various traditional approaches for ToF data improvement have been proposed, machine learning techniques have seldom been applied to this task, mostly due to the limited availability of real world training data with depth ground truth. In this paper, we avoid to rely on labeled real data in the learning framework. A Coarse-Fine CNN, able to exploit multi-frequency ToF data for MPI correction, is trained on synthetic data with ground truth in a supervised way. In parallel, an adversarial learning strategy, based on the Generative Adversarial Networks (GAN) framework, is used to perform an unsupervised pixel-level domain adaptation from synthetic to real world data, exploiting unlabeled real world acquisitions. Experimental results demonstrate that the proposed approach is able to effectively denoise real world data and to outperform state-of-the-art techniques.

Block based recursive search for motion estimation is already suitable for real time applications thanks to its wide utilization, but the increase of resolution in modern displays and the need for real time software implementations demand for faster approaches. In this paper a new motion estimation procedure that has been developed for a parallel implementation is presented. The parallel implementation was achieved by avoiding the spatial recursion which is typical for the recursive search systems, while the convergence can still be obtained by means of the temporal recursion only. In particular the use of a binarized cross correlation as a matching cost criteria instead of the standard SAD permits an improved motion vector field and a faster convergence without increasing the computational cost. The experimental results show a similar vector field estimation in comparison to the standard SAD based recursive search method with the considerable advantage of a possible parallel implementation.

Among all the techniques for 3D acquisition, stereo vision systems are the most common. More recently, Time-of-Flight (ToF) range cameras have been introduced. They allow real-time depth estimation in conditions where stereo does not work well. Unfortunately, ToF sensors still have a limited resolution (e.g., 200 × 200 pixels). The goal of this paper is to combine the information from the ToF with one or two standard cameras, in order to obtain a highresolution depth image. To this end, we propose a new bilateral filter for depth maps up-sampling, which exploits the additional information provided by a high-resolution color image. Moreover, we present an entire framework for the super-resolution of a real ToF depth map with a single camera color image. The algorithm allows to enhance the resolution of the ToF camera depth image up to 1920 × 1080 pixels.

In this paper we propose an effective de-ringing method for DCT compressed images and videos. This method performs a subband decomposition provided by a translation invariant wavelet and processes only the high level subbands by means of a local adaptive equalization around the edges of the image. This makes the procedure suitable for both texture and flat areas, and moreover for different compression ratios. The benefits of the algorithm are then shown with objective, based on PSNR and SSIM metrics, and subjective evaluations.