Research On Dynamic Texture Modeling And Synthesis Under Image Reconstruction And Kernel Learning

Dynamic textures(DTs) have been successfully modeled using linear dynamic systems(LDS). This method is flexible, requires less memory, and allows powerful editing. However, since the over-simplified noise-driven LDS model is not always stable or oscillates, it cannot synthesize long and visually pleasing DT video sequences.In order to solve the under fit problem caused by the linear observation model in LDS, we firstly propose a new dynamic texture synthesis framework via creatively fitting the basic LDS with a newly developed image reconstruction technique to enhance texture details while maintaining the temporal coherence of the reconstructed texture images. Experiments on standard dynamic texture databases demonstrate that our method exhibits superior performance on synthesizing dynamic textures.Furthermore, we show that the widely-used LDS model can be approximated using a principal component regression(PCR) model with the main advantage of simplicity. To capture the nonlinearity of training frames, we extend traditional PCR to its kernelized version and introduce kernel principal component regression(KPCR) to model and synthesize DTs. To meet the demand for online learning, we remove the standard state model and directly apply the quantized kernel least mean squares(QKLMS) algorithm in signal processing domain to approximate the performance achieved with KPCR. We term this improvement kernel adaptive dynamic texture synthesis(KADTS), which also has the benefits of computational and memory efficiency. These advantages make KADTS ideally suited for real world applications, since the majority of electronic devices, including cell phones and laptops, suffer from limited memory and real-time constraints. We demonstrate, via both theoretical and experimental analyses, the connections between DT synthesis using KADTS and KPCR with regularization network theory. We also show the superiority of our proposed algorithms for DT synthesis compared to other dynamic system based benchmarks.