A weighted compressive sensing method for multi-view images

Compressive Sensing (CS) is a stable and robust technique that allows for the sub-sampling of data at a given data rate: `compressive sampling´ or `compressive sensing´ at rates smaller than the Nyquist sampling rate. while theoretical studies have demonstrated the stability of CS, specific examples of successful and practical applications remain elusive. In this paper, we apply multi-view images obtained from multiple visual sensor nodes to CS, where the measurement side emphasizes `important´ CS data in view of 3D reconstruction. Due to the spatial proximity of cameras, the obtained images have high correlations. However, cameras cannot collaborate in image acquisition. Therefore, compression should be performed locally at each camera and reconstruction is executed jointly to consider dependencies in the acquired data.