Convolutional Neural Networks (CNN) are widely used for image classification and have achieved significantly accurate performance in the last decade. However, they require computationally intensive operations for embedded applications. In recent years, FPGA-based CNN accelerators have been proposed to improve energy efficiency and throughput. While dynamic partial reconfiguration (DPR) is increasingly used in CNN accelerators, the performance of dynamically reconfigurable accelerators is usually lower than the performance of pure static FPGA designs. This work presents a dynamically reconfigurable CNN accelerator architecture that does not sacrifice throughput performance or classification accuracy. The proposed accelerator is composed of reconfigurable macroblocks and dynamically utilizes the device resources according to model parameters. Moreover, we devise a novel approach, to the best of our knowledge, to hide the computations of the pooling layers inside the convolutional layers, thereby further improving throughput. Using the proposed architecture and DPR, different CNN architectures can be realized on the same FPGA with optimized throughput and accuracy. The proposed architecture is evaluated by implementing two different LeNet CNN models trained by different datasets and classifying different classes. Experimental results show that the implemented design achieves higher throughput than current LeNet FPGA accelerators.

In this study, we propose a novel single image Bayesian super-resolution (SR) algorithm where the hyperspectral image (HSI) is the only source of information. The main contribution of the proposed approach is to convert the ill-posed SR reconstruction (SRR) problem in the spectral domain to a quadratic optimization problem in the abundance map domain. In order to do so, Markov Random Field (MRF) based energy minimization approach is proposed and proved that the solution is quadratic. The proposed approach consists of five main steps. First, the number of endmembers in the scene is determined using virtual dimensionality. Second, the endmembers and their low resolution abundance maps are computed using simplex identification via the splitted augmented Lagrangian (SISAL) and fully constrained least squares (FCLS) algorithms. Third, high resolution (HR) abundance maps are obtained using our proposed maximum a posteriori (MAP) based energy function. This energy function is minimized subject to smoothness, unity and boundary constraints. Fourth, the HR abundance maps are further enhanced with texture preserving methods. Finally, HR HSI is reconstructed using the extracted endmembers and the enhanced abundance maps. The proposed method is tested on three real HSI datasets; namely the Cave, Harvard and Hyperspectral Remote Sensing Scenes (HRSS) and compared to state-of-the-art alternative methods using peak signal to noise ratio, structural similarity, spectral angle mapper and relative dimensionless global error in synthesis metrics. It is shown that the proposed method outperforms the state of the art methods in terms of quality while preserving the spectral consistency.

Super-resolution Reconstruction (SRR) is technique to increase the spatial resolution of images. It is especially useful for hyperspectral images (HSI), which have good spectral resolution but low spatial resolution. In this study, we propose an improvement to our previous work and present a novel MAP-MRF (maximum a posteriori-Markov random Fields) based approach for the SRR of HSI. The key point of our approach is to find the abundance maps of an HSI and perform SRR on the abundance maps using MRF based energy minimization, without needing any other additional source of information. In order to do so, first, PCA is used to determine the endmembers. Second, SISAL and fully constraint least squares (FCLS) are used to estimate the abundance maps. Third, in order to find the high resolution abundance maps, the ill-posed inverse SRR problem for abundances is regularized with a MAP-MRF based approach. The MAP-MRF formulation is restricted with the constraints which are specific to the abundances. Using the non-linear programming (NLP) techniques, the convex MAP formulation is minimized and High Resolution (HR) abundance maps are obtained. Then, these maps are used to construct the HR HSI. This improved SRR method is verified on real data sets, and quantitative performance comparison is achieved using PSNR, SSIM and PSNR metrics. Our results indicate that this improved method gives very close results to the original high resolution images, keeps the spectral consistency, and performs better than the compared algorithms.