Percorrer por autor "Filipe, Jose N."
A mostrar 1 - 4 de 4
Resultados por página
Opções de ordenação
- Complexity Estimation for Load Balancing of 360-Degree Intra Versatile Video CodingPublication . Filipe, Jose N.; Monteiro Carreira, João Filipe; Tavora, Luis M. N.; Faria, Sergio; Navarro, Antonio; Assuncao, Pedro A. A.The ever increasing demand for image and video content poses new requirements to support higher resolutions and richer representation formats, creating new challenges in coding algorithms. The forthcoming Versatile Video Coding (VVC) standard aims to increase the coding efficiency of existing algorithms and it is particularly suitable for Ultra-High Definition (UHD) resolutions and 360° video. However, since coding efficiency gains are obtained at the cost of increased complexity, fast computational approaches are needed to cope with realtime requirements, such as parallel processing. Thus, this work presents a contribution towards efficient parallel encoding of 360° video, based on coding complexity estimation and nonuniform data-level splitting (slice-based) for load balancing across multiple processors. A machine learning approach is proposed to estimate the complexity of intra coding VVC, using uncorrelated features, obtained through Principal Component Analysis (PCA) and Extremely Randomised Trees (ERT). Then, a complexity-balanced slice partition is devised, taking advantage of the clustered complexity inherent to Equirectangular Projection (ERP). It is shown that coding complexity is estimated with an accuracy of 92.25%, and the encoding time is reduced by 8.50%, when compared to the case where the 360° frames are evenly split.
- Evaluation of Focus Metrics in Extended Depth-of-field ReconstructionPublication . Filipe, Jose N.; Távora, Luis M.N.; Assunção, Pedro A. A.; Fonseca-Pinto, Rui; Faria, Sérgio M. M. deThe performance of focus metrics in the evaluation of refocused images from light fields is investigated in this paper. To this aim, the paper presents a comprehensive study on the performance of a large set of different focus metrics (34 in total) in the evaluation of patch-based extended depth-of-field reconstructed images (e.g., all-in-focus). The new findings of this work demonstrate that optimal reconstruction of extended depth-of-field images, from light fields captured with focused plenoptic cameras, is not consistent with the focus level figures computed by available focus metrics. The results show that a higher focus level, as given by such computational methods, does not actually correspond to better all-in-focus images, thus indicating that currently available focus metrics are not suitable for evaluating the quality of extended depth-of-field reconstruction. The results obtained through subjective evaluation confirm that objective focus metrics fail to indicate the best focused images. Finally, the paper presents recommendations to adapt existing metrics to extended depth-of-field reconstruction in plenoptic imaging.
- Improved Patch-based View Rendering for Focused Plenoptic Cameras with Extended Depth-of-FieldPublication . Filipe, Jose N.; Assuncao, Pedro A. A.; Tavora, Luis M. N.; Fonseca-Pinto, Rui; Thomaz, Lucas A.; Faria, Sergio M. M.This paper proposes a patch-based framework to improve view rendering from light fields captured by focused plenoptic cameras with extended depth-of-field. A three-step approach is performed to minimise discontinuities across neighbouring patches and to maximise the information captured from different types of micro-lens. Firstly, the useful patch size of each micro-image is determined to limit the contribution of low light intensity regions into the rendered images. Secondly, anisotropic inpainting is applied to smooth the discontinuities between neighbouring patches. Thirdly, a multi-focus image fusion algorithm is used to efficiently combine the information captured by the different types of micro-lens comprising plenoptic cameras with extended depth-of-field. Significant improvements are obtained in the rendered views, by reducing visible discontinuities between adjacent micro-images, while at the same time increasing the average sharpness of image content by 20%, measured by the Energy of Laplacian (i.e., a focus metric), in comparison to the conventional rendering approach, which always selects the best focused lens type.
- Tree-Based Ensemble Methods for Complexity Reduction of VVC Intra CodingPublication . Filipe, Jose N.; Carreira, J.; Tavora, Luis M. N.; Faria, Sergio M. M. de; Navarro, Antonio; Assuncao, Pedro A. A.The rise of new applications and emerging services required ultra-high definition (UHD) led to the development of a new video compression standard, name Versatile Video Coding (VVC). In comparison with High Efficiency Video Coding, the new standard increases the coding efficiency, at the expense of greatly increasing the computational complexity of the encoding process, especially in UHD and 360°video. To overcome this issue, this paper presents a novel method to reduce the computational complexity of intra-coded 360°video in Equirectangular Projection format, based on 3 tree-based ensemble models to predict the maximum partition depth that should be used for intra-coding of the complex nested data structures (Quad Tree, Binary Tree and Ternary Tree) used in the VVC standard. Two different tree-based ensemble methods are studied, namely Random Forest and Extremely Randomised Trees. The experimental results achieve, on average, complexity reductions of 56.25% and 57.45% for Random Forests and Extremely Randomised Trees methods, with small loss of coding efficiency of 1.37 % and 1.46 %, respectively.