Compressed Sensing for Remote Imaging in Aerial and Terrestrial Surveillance
  1. Compressive Video Sensing & Classification
  2. GPU accelerated CVS
  3. Compressed Gated Range Sensing (CGRS)
  4. Compressed Sensing & Sparsity in Imaging
  5. Compressed Sensing in Localization
  6. Fundamentals of CS Acquisition and Reconstruction

The M-JPEG-based encoder that was introduced by G. Tzagkarakis et al. [7] consists of several modules that handle different operations of the video encoding process. We investigated more effective optimization approaches that minimize the execution time of the CS measurement acquisition process. In specific, we introduce a faster multicore implementation and also a GPU porting of the CS module that outperforms all optimization techniques that we have tried so far. In order to demonstrate the performance of the different approaches, we conducted three experiments: Execution Time vs Frame size, Execution time vs Block Size and Execution Time vs Sampling Ratio.

Processor Details
Host (PC) Intel Core i5-3230M CPU @ 2.6 GHz
4 Logical Cores
8 Gb RAM
Device (GPU) NVidia NVS 5200M
96 Cores
1 Gb Memory
Compute Capability: 2.1
Parameter Value
Video Sequence Akiyo YUV QCIF (176 x 144) Monochrome
Frames Processed 50 (55 with GOP correction algorithm)
GOP size 6
Number of processed GOPs 9
Measurement Matrix ORTH

Figure 1: Execution Time vs Frame Size

Figure 1: Execution Time vs Frame Size

Figure 3: Execution Time vs Block Size

Figure 4: Speedup vs CS Block Size
ORION members involved: Kostas Lekkas, George Tzagkarakis.