Publicação
Scalable light field representation and coding
| Resumo: | This Thesis aims to advance the state-of-the-art in light field representation and coding. In this context, proposals to improve functionalities like light field random access and scalability are also presented. As the light field representation constrains the coding approach to be used, several light field coding techniques to exploit the inherent characteristics of the most popular types of light field representations are proposed and studied, which are normally based on micro-images or sub-aperture-images. To encode micro-images, two solutions are proposed, aiming to exploit the redundancy between neighboring micro-images using a high order prediction model, where the model parameters are either explicitly transmitted or inferred at the decoder, respectively. In both cases, the proposed solutions are able to outperform low order prediction solutions. To encode sub-aperture-images, an HEVC-based solution that exploits their inherent intra and inter redundancies is proposed. In this case, the light field image is encoded as a pseudo video sequence, where the scanning order is signaled, allowing the encoder and decoder to optimize the reference picture lists to improve coding efficiency. A novel hybrid light field representation coding approach is also proposed, by exploiting the combined use of both micro-image and sub-aperture-image representation types, instead of using each representation individually. In order to aid the fast deployment of the light field technology, this Thesis also proposes scalable coding and representation approaches that enable adequate compatibility with legacy displays (e.g., 2D, stereoscopic or multiview) and with future light field displays, while maintaining high coding efficiency. Additionally, viewpoint random access, allowing to improve the light field navigation and to reduce the decoding delay, is also enabled with a flexible trade-off between coding efficiency and viewpoint random access. |
|---|---|
| Autores principais: | Monteiro, Ricardo Jorge Santos |
| Assunto: | Light field coding Light field representation Viewpoint scalability Viewpoint random access Campo de luz -- Field of light Codificação de campos de luz Representação de campos de luz Escalabilidade de ponto de vista Acesso aleatório de ponto de vista Processamento de imagens Codificação -- Codification Imagem digital |
| Ano: | 2020 |
| País: | Portugal |
| Tipo de documento: | tese de doutoramento |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | ISCTE |
| Idioma: | inglês |
| Origem: | Repositório ISCTE |
| Resumo: | This Thesis aims to advance the state-of-the-art in light field representation and coding. In this context, proposals to improve functionalities like light field random access and scalability are also presented. As the light field representation constrains the coding approach to be used, several light field coding techniques to exploit the inherent characteristics of the most popular types of light field representations are proposed and studied, which are normally based on micro-images or sub-aperture-images. To encode micro-images, two solutions are proposed, aiming to exploit the redundancy between neighboring micro-images using a high order prediction model, where the model parameters are either explicitly transmitted or inferred at the decoder, respectively. In both cases, the proposed solutions are able to outperform low order prediction solutions. To encode sub-aperture-images, an HEVC-based solution that exploits their inherent intra and inter redundancies is proposed. In this case, the light field image is encoded as a pseudo video sequence, where the scanning order is signaled, allowing the encoder and decoder to optimize the reference picture lists to improve coding efficiency. A novel hybrid light field representation coding approach is also proposed, by exploiting the combined use of both micro-image and sub-aperture-image representation types, instead of using each representation individually. In order to aid the fast deployment of the light field technology, this Thesis also proposes scalable coding and representation approaches that enable adequate compatibility with legacy displays (e.g., 2D, stereoscopic or multiview) and with future light field displays, while maintaining high coding efficiency. Additionally, viewpoint random access, allowing to improve the light field navigation and to reduce the decoding delay, is also enabled with a flexible trade-off between coding efficiency and viewpoint random access. |
|---|