Adaptive gradient-based analog hardware architecture for 2D under-sampled signals reconstruction

The paper proposes an analog hardware solution for the implementation of two-dimensional gradient-based algorithm. The algorithm employs the discrete cosine transform and performs missing samples reconstruction by using the compressive sensing principles. Although there is a number of algorithms for solving two-dimensional compressive sensing problems, for many of them a real-time application is a challenging task. Therefore, this paper observes an algorithm whose real-time application has relatively low complexity. Also, the reconstruction accuracy is comparable to the commonly used compressive sensing algorithms. The algorithm is observed within the several parts. The implementation of each part is considered in details, with provided discussion on the computational complexity of each part.