Nonuniform sampling with adaptive expectancy based on local variance

A new trend in sensor array architectures is to provide compact implementations based on alternative acquisitions and sampling. In particular, with the recent rise of Compressive Sensing (CS), multiple sensing schemes have been developed. However, for the moment, CS reconstruction techniques take a relatively long time to properly converge. Therefore, it limits the sensor resolution and potential applications. On the other hand, it generally involves complex structures and circuitries at the sensor side. This work proposes an acquisition chain performing an adaptive sensing of pseudo-randomly selected samples. This specific nonuniform sampling scheme allows to parallelize and simplify the acquisition thanks to a compact design based on sigma delta converters and cellular automata. Previous works show that compared to state of the art and without an important image degradation, dedicated reconstructions to this specific sampling can considerably reduce the overall computation time.