Plug-and-Play Adaptive Block Compressive Sensing using Edge-Detection on the Compressed Domain

Current adaptive sensing methods impose higher hardware and time costs, needing two sensors and dual reconstructions.

To address this, this paper integrates edge detection on sampled data for adaptive sampling. The edge detection outcome dictates the sampling rate for each block.
Additionally, a strategy is proposed to regulate the overall sampling rate across the entire image.
This adaptive method, processed on the sampled data, requires only a single sensor and reconstruction, thus compatible with any sampling matrix, and reduces time and hardware costs.

Experimental comparisons against traditional and learning-based techniques consistently reveal superior results.