Tesis doctoral de Piedad Brox Jimenez
The main goal of this research work has been to develop and to implement in hardware fuzzy logic-based algorithms capable of improving state-of-the-art approaches when solving image interpolation problems, particularly video de-interlacing problems. interlacing was introduced in the early television transmission systems to halve the video bandwidth without degrading the image quality exceedingly. Its success was initially supported by the fact that cathode ray tubes (crts) can display interlaced video directly as a result of the persistence of the phosphor of crts. Furthermore, some of the current hdtv standards use an interlaced scanning but a progressive scanning is required by the majority of modern display devices. Hence, the development of algorithms that convert an interlaced video signal into a progressive one is an active research field. starting from motion-adaptive de-interlacing algorithms reported in the literature as a good midpoint between simple linear de-interlacing algorithms and complex motion-compensated ones, our first contribution has been the development of a fuzzy motion-adaptive algorithm that uses a simple convolution to measure the motion at each pixel and employs 3 simple fuzzy rules to adapt interpolation to motion. taking into account the especial sensitivity of human visual system to the edges in images, our second contribution has been the development of a fuzzy edge-dependent spatial interpolator that evaluates 5 potential edge directions and employs 6 simple fuzzy rules to adapt interpolation to edges. since interlaced material could come from many sources (film, video and hybrid), detection of pattern repetition is becoming very valuable. In this sense, our third contribution has been the development of a fuzzy picture-repetition-dependent temporal interpolator that uses a simple convolution to measure the dissimilarity between consecutive fields and employs 2 simple fuzzy rules to adapt interpolation to repetition. the final algorithm presented not only performs better than other algorithms of similar complexity (such as vertico-temporal and median-based approaches) but also improves the results obtained by algorithms of much greater complexity (such as several motion-compensated ones). taking into account the flexibility and low cost of current fpgas, hardware implementations developed through this research have been done on these programmable devices. A computer-aided design methodology has been employed to cross the way easily and rapidly from high-level algorithm development to synthesis and experimental validation. At algorithmic level, matlab and its image processing toolbox have been employed to develop image processing algorithms. Xfuzzy 3 and its xfsl tool have facilitated tuning the parameters of the different fuzzy rule bases proposed. Xsg from xilinx has paved the way to develop efficient hardware-description-language (hdl) definitions from algorithmic level. Finally, cad environments to easily develop fpga designs from hdl descriptions (in particular ise and edk from xilinx) have been exploited to develop the demonstrators that have validated experimentally our proposals.
Datos académicos de la tesis doctoral «Fuzzy logic-based algorithm for video de-interlacing and its hardware implementation«
- Título de la tesis: Fuzzy logic-based algorithm for video de-interlacing and its hardware implementation
- Autor: Piedad Brox Jimenez
- Universidad: Sevilla
- Fecha de lectura de la tesis: 08/05/2009
Dirección y tribunal
- Director de la tesis
- Santiago Sánchez Solano
- Tribunal
- Presidente del tribunal: José Luis Huertas diaz
- mike Nachtegael (vocal)
- gerard De haan (vocal)
- claudio Moraga roco (vocal)