Gaussian Splatting (GS) offers a promising alternative to Neural Radiance Fields (NeRF) for real-time 3D scene rendering. Using a set of 3D Gaussians to represent complex geometry and appearance, GS achieves faster rendering times and reduced memory consumption compared to the neural network approach used in NeRF. However, quality assessment of GS-generated static content is not yet explored in-depth. This paper describes a subjective quality assessment study that aims to evaluate synthesized videos obtained with several static GS state-of-the-art methods. The methods were applied to diverse visual scenes, covering both 360° and forward-facing (FF) camera trajectories. Moreover, the performance of 18 objective quality metrics was analyzed using the scores resulting from the subjective study, providing insights into their strengths, limitations, and alignment with human perception. All videos and scores are made available providing a comprehensive database that can be used as benchmark on GS view synthesis and objective quality metrics.

Neural Radiance Fields (NeRF) have revolutionized the field of 3D visual representation by enabling highly realistic and detailed scene reconstructions from a sparse set of images. NeRF uses a volumetric functional representation that maps 3D points to their corresponding colors and opacities, enabling photorealistic view synthesis from arbitrary viewpoints. Despite its advancements, the efficient streaming of NeRF content remains a significant challenge due to the large amount of data involved. This paper investigates the rate-distortion performance of two NeRF streaming strategies: pixel-based and neural network (NN) parameter-based streaming. While in the former, images are coded and then transmitted throughout the network, in the latter, the respective NeRF model parameters are coded and transmitted instead. This work also highlights the trade-offs in complexity and performance, demonstrating that the NN parameter-based strategy generally offers superior efficiency, making it suitable for one-to-many streaming scenarios.

To render a spherical (360° or omnidirectional) image on planar displays, a 2D image - called as viewport - must be obtained by projecting a sphere region on a plane, according to the user's viewing direction and a predefined field of view (FoV). However, any sphere to plan projection introduces geometric distortions, such as object stretching and/or bending of straight lines, which intensity increases with the considered FoV. In this paper, a fully automatic content-aware projection is proposed, aiming to reduce the geometric distortions when high FoVs are used. This new projection is based on the Pannini projection, whose parameters are firstly globally optimized according to the image content, followed by a local conformality improvement of relevant viewport objects. A crowdsourcing subjective test showed that the proposed projection is the most preferred solution among the considered state-of-the-art sphere to plan projections, producing viewports with a more pleasant visual quality.

This short paper proposes a new database - NeRF-QA - containing 48 videos synthesized with seven NeRF based methods, together with their perceived quality scores, resulting from subjective assessment tests; for the videos selection, both real and synthetic, 360 degrees scenes were considered. This database will allow to evaluate the suitability, to NeRF based synthesized views, of existing objective quality metrics, and also the development of new quality metrics.

Detecting and diagnosing the root cause of failures in mobile networks is an increasingly demanding and time-consuming task, given its technological growing complexity. This paper focuses on predicting and diagnosing low User Downlink (DL) Average Throughput situations, using supervised learning and the Tree Shapley Additive Explanations (SHAP) method. To fulfill this objective, Boosting classification models are used to predict a failure/non-failure binary label. The influence of each counter on the overall model’s predictive performance is performed based on the TreeSHAP method. From the implementation of this technique, it is possible to identify the main causes of low throughput, based on the analysis of the most critical counters in fault detection. Furthermore, from the identification of these counters, it is possible to define a system for diagnosing the most probable throughput degradation cause. The described methodology allowed not only to identify and quantify low throughput situations in a live network due to the occurrence of misadjusted configuration parameters, radio problems and network capacity problems, but also to outline a process for solving them.

With the increase in consumption of multimedia content through mobile devices (e.g., smartphones), it is crucial to find new ways of optimizing current and future wireless networks and to continuously give users a better Quality of Experience (QoE) when accessing that content. To achieve this goal, it is necessary to provide Mobile Network Operator (MNO) with real time QoE monitoring for multimedia services (e.g., video streaming, web browsing), enabling a fast network optimization and an effective resource management. This paper proposes a new QoE prediction model for video streaming services over 4G networks, using layer 1 (i.e., Physical Layer) key performance indicators (KPIs). The model estimates the service Mean Opinion Score (MOS) based on a Machine Learning (ML) algorithm, and using real MNO drive test (DT) data, where both application layer and layer 1 metrics are available. From the several considered ML algorithms, the Gradient Tree Boosting (GTB) showed the best performance, achieving a Pearson correlation of 78.9%, a Spearman correlation of 66.8% and a Mean Squared Error (MSE) of 0.114, on a test set with 901 examples. Finally, the proposed model was tested with new DT data together with the network's configuration. With the use case results, QoE predictions were analyzed according to the context in which the session was established, the radio transmission environment and radio channel quality indicators.

To visualize the spherical (360°) content on 2D planar displays, a selected part of the sphere must be projected into a 2D plane. The resulting 2D area is called viewport and is created according some fixed and limited field of view, usually much less than 360°. To generate the viewport, two geometric projections are usually employed: rectilinear and stereographic. For both cases, this process leads to geometric distortions in the viewport image which, depending on the content, may be visually unpleasant to users. Rectilinear projection stretches the image borders while the stereographic projection bends straight lines. In this paper, the viewport rendering problem is tackled with the formulation of a generalized perspective projection (which includes many other popular projection, besides perspective and stereographic) and a perceptual optimization of its key parameter: the projection center. The optimization process is performed according to the viewport characteristics, namely with perceptually important features that characterize the amount of line distortion and stretching of salient regions. The aim is to find an optimal, content dependent, projection center that minimizes the content distortions and thus provides better quality of experience (QoE) in the visualization of spherical images. Experimental results on several 360° images show that the proposed optimized projection achieves better results than the benchmark projections both quantitatively and qualitatively.

Performing an efficient management of mobile networks has become an increasingly difficult task, as these networks are becoming more complex, which results in more data to evaluate and optimize. This paper is focused on the performance evaluation of a Long Term Evolution (LTE) network using unsupervised learning techniques. The main objective is to detect groups of cells that show similar performances and, consequently, to identify the groups that perform below the desired threshold. Additionally, this work also aims to identify which cell configurations are associated with the best performance. In order to fulfill the first objective, a methodology based on the application of clustering algorithms to features extracted from Key Performance Indicators (KPIs) was developed. Regarding the second objective, statistical testing was applied to evaluate the dependence between the cells configuration and the respective cells performance. It was verified that there is not a significant difference in the obtained results using different clustering algorithms. In the majority of the cases, only two groups of cells were identified: one group consisting essentially on the cells with the best performance and the other group containing the worst performing cells. As far as the connection between configuration data and performance data is concerned, only one influencing case, referring to a parameter associated with the cell subscription capacity, was detected.

In this paper, a novel quality metric to evaluate depth-based synthesized views is proposed. This metric relies on a hybrid approach that uses features extracted in different phases of the image synthesis procedure, namely from the bitstream, from intermediate data produced during the synthesis process, and from the final synthesized view; these features are then combined using support vector regression. A new data set of synthesized images, with compression and rendering artifacts, was built and used to develop and assess the proposed metric. The metric performance is compared with conventional full-reference two-dimensional image quality assessment metrics and with quality assessment metrics developed specifically for synthesized images. The experimental results showed that the proposed solution outperforms the considered benchmark metrics, being able to predict the subjective quality scores of the synthesized images with a Pearson correlation coefficient close to 0.9.

This paper presents a novel Quality of Experience (QoE) prediction model, for voice and web browsing based on real Universal Mobile Telecommunications System (UMTS) and Long Term Evolution (LTE) data. It estimates the user perceived quality, in a Mean Opinion Score (MOS) scale, by evaluating several Radio Frequency (RF) channel measurements and Quality of Service (QoS) metrics. Real Drive Test (DT) data and MOS measurements were used as reference data, in order to produce a new QoE prediction model, using machine learning techniques. The Support Vector Regression (SVR) algorithm was used to map the QoS metrics into MOS. The new developed model enables the application of QoE as a more realistic network optimization criteria. The QoE model for voice calls presented a Root Mean Square Error (RMSE) of 11% and a correlation of 62%, when comparing the predicted MOS to the one that was measured. The web browsing model showed an higher correlation (of 92%) and a lower RMSE (of 10%).

In the last years, the video content consumed by mobile users has increased exponentially. Since mobile network capacity cannot be increased as fast as required, it is crucial to develop intelligent schedulers that allocate radio resources very efficiently and are able to provide a high Quality of Experience (QoE) to most of the users. This paper proposes a new and effective scheduling solution—the Maximum Buffer Filling (MBF) algorithm—which aims to increase the number of satisfied users in video streaming services provided by wireless networks. The MBF algorithm uses the current buffer level at the client side and the radio channel conditions, which are reported to the network by the client, as well as the bitrate of the requested video segment. The proposed scheduling strategy can also fulfill different satisfaction criteria, since it can be tuned to maximize the numbers of users with high QoE levels or to minimize the number of users with low QoE levels. A simulation framework was developed, considering a Long Term Evolution (LTE) scenario, in order to assess the performance of the proposed scheduling scheme and to compare it with other well‐known scheduling solutions. The results show the superior performance achieved by the proposed technique, in terms of the number of satisfied and unsatisfied users.

Omnidirectional, also referred to as 360o, visual content provides an immersive experience since it allows users to view a visual scene from different directions. The overall content typically covers a full sphere, and omnidirectional videos or images are processed to obtain a projection on a 2D plane of a fraction of the sphere (aka viewport), which is shown to the user. Therefore, users can look around freely and only a restricted field of view (FoV) of the entire content is shown, according to the selected viewing direction. To maximize the immersion and engagement that omnidirectional content can provide, viewports must be generated and displayed with a large FoV. However, when large FoVs are used, geometric distortions (e.g., shape and scale deformations) can be present. In this paper, the perceptual impact of the perspective projection for viewport rendering of 360oimages is evaluated. This study is based on subjective experiments conducted at Instituto Superior Técnico (IST), aiming to assess the perceptual impact of the rendering projection parameters, such as FoV and centre of projection, as well as the influence of the image content.

In this paper the spatial distribution of the channel capacity of Multiple Input Multiple Output (MIMO) wireless systems within a microcell scenario is studied. Based on the capacity location dependency, made clear by realistic simulation results, a segmentation of the cell area into high and low capacity regions is proposed. The Base Station (BS) and Mobile Stations (MSs) inter-antenna spacing impact on these regions is also investigated. The results suggest the most efficient ways to enlarge the high capacity region, providing a framework to optimize the design and deployment of MIMO systems.

This paper proposes a no-reference quality assessment metric for digital video subject to H.264/advanced video coding encoding. The proposed metric comprises two main steps: coding error estimation and perceptual weighting of this error. Error estimates are computed in the transform domain, assuming that discrete cosine transform (DCT) coefficients are corrupted by quantization noise. The DCT coefficient distributions are modeled using Cauchy or Laplace probability density functions, whose parameterization is performed using the quantized coefficient data and quantization steps. Parameter estimation is based on a maximum-likelihood estimation method combined with linear prediction. The linear prediction scheme takes advantage of the correlation between parameter values at neighbor DCT spatial frequencies. As for the perceptual weighting module, it is based on a spatiotemporal contrast sensitivity function applied to the DCT domain that compensates image plane movement by considering the movements of the human eye, namely smooth pursuit, natural drift, and saccadic movements. The video related inputs for the perceptual model are the motion vectors and the frame rate, which are also extracted from the encoded video. Subjective video quality assessment tests have been carried out in order to validate the results of the metric. A set of 11 video sequences, spanning a wide range of content, have been encoded at different bitrates and the outcome was subject to quality evaluation. Results show that the quality scores computed by the proposed algorithm are well correlated with the mean opinion scores associated to the subjective assessment.

In multiple input multiple output (MIMO) systems the antenna array configuration in the base station (BS) and mobile station (MS) has a large influence on the available channel capacity. In this paper, we first introduce a new frequency selective (FS) MIMO framework for macro-cells in a realistic urban environment. Next, MIMO configuration characteristics are investigated in order to maximize capacity, mainly the number of antennas and inter-antenna spacing. Channel and capacity simulation results are presented for the city of Lisbon, Portugal, using different antenna configurations. Two power allocations schemes are considered, uniform distribution and FS spatial water-filling. The results suggest optimized MIMO configurations, considering the antenna array size limitations, at the MS side.

In this paper a realistic model that is an extension of COST 273 channel model, introducing terrain and clutter information into cluster generation, is presented. Our model is based on a geometrical/stochastic approach, and allows a realistic treatment of both cluster positions and dynamics. This is done by means of visibility regions. Using terrain and clutter information, the cluster and visibility region generating PDFs are derived. Some simulation results over a GIS tool for the city of Lisbon, Portugal, are presented.

The paper introduces an algorithm that converts dynamically GIS terrain and clutter information into the flat terrain input parameterization needed for COST 231 Walflsch-Ikegami model. Under this algorithm, a new method for average building height calculation is presented, using Epstein-Peterson's multiple knife-edge diffraction method. Additionally, it is introduced a new realistic approach for line-of-sight stochastic modeling. The mobile station (MS) paths are generated using a new mobility model previously validated for the city of Lisbon, Portugal.

This paper describes a watermark-based technique that scores the quality of an image by blind estimation of its PSNR. Watermark embedding is performed in the block-based DCT domain using a non-uniform quantization scheme that accounts for human vision characteristics. At the extraction, the square distances between received coefficients and nearest quantization points are computed, giving an estimative for the image mean square error (and consequently, its PSNR). Since these distances may be underestimated if distortion is greater than the quantization step used for watermark embedding, it is also proposed the use of watermark extraction bit error rate (together with image statistics) as means of PSNR estimation weighting. Results show that PSNR estimation closely follows the true PSNR, for images subject to JPEG and JPEG2000 compression.

This paper is focused on image authentication, as the process of evaluating the integrity of image contents relative to the original picture and of being able to detect, in an automatic way, malevolent image modifications. The paper begins with a description of a general framework for content based authentication of images and video. Then, a specific method is proposed. It relies on image edges and tries to tackle the problem of image/video integrity from a semantic, high-level point of view. Experimental results are presented for well known image and video sequences.

In standard DCT coding schemes like MPEG, the sequences compression is achieved by motion compensation, transformation, quantization, and entropy coding. In this paper, we have followed the same path by adapting to the image signal the elements of the coding scheme. The motion compensation is achieved by a block-matching method, where the size of the blocks is adapted to the signal. Great attention has been paid to the relevance of the motion field. Combined with the motion compensation, the two fields of each frame are merged, taking into account the measured motion vectors, to compose a pseudo-progressive frame. The encoding is applied to this `motion-compensated progressive' frame. A wavelet decomposition is then applied on each (inter or intra) frame. Such a transform, intrinsically owning linear- phase and perfect reconstruction properties, has been optimized for maximizing a perceptually weighted coding gain. The wavelet coefficients are thereafter vector-quantized, in order to reach the maximum perceptual SNR : frequency weighting is taken into account. The relevance of the measured vector field allows a precise spatio-temporal quantization optimization. The vectors are entropy coded taking into account the remaining inter-band dependence, by an adapted entropy code. Results obtained from 1 Mbit/s to 8 Mbit/s are shown for moving sequences at the conference.