Figure 5. Balanced single-pixel video imaging.

We show multiple frames of a live video taken with our single-pixel camera. In the top row, we show the frames acquired without using subsampling techniques. As the images consist of 64 × 64 pixels, the 4096 measurements required to take a frame were performed in roughly 200 ms. In the middle row we show frames of the same scene acquired by subsampling the scene. In this case, we used 820 patterns, which correspond to a 20% compression ratio. Each frame was acquired in 40 ms. In the lower row we show a CS reconstruction of the frames obtained before. In this case, the quality drop is almost negligible and we still can record a video at 25 Hz. The numbers shown correspond to the frame number inside the video.