Compressive Imaging: A New Framework for Computational Image Processing

November 7, 2005
  • Image processing
  • 68U10
Imaging sensors, hardware, and algorithms are under increasing pressure to accommodate ever larger and higher-dimensional data sets; ever faster capture, sampling, and processing rates; ever lower power consumption; communication over ever more difficult channels; and radically new sensing modalities. Fortunately, over the past few decades, there has been an enormous increase in computational power and data storage capacity, which provides a new angle to tackle these challenges. We could be on the verge of moving from a "digital signal processing" (DSP) paradigm, where analog signals (including light fields) are sampled periodically to create their digital counterparts for processing, to a "computational signal processing" (CSP) paradigm, where analog signals are converted directly to any of a number of intermediate, "condensed" representations for processing using optimization techniques. At the foundation of CSP lie new uncertainty principles that generalize Heisenberg's between the time and frequency domains and the concept of compressibility. As an example of CSP, I will overview "Compressive Imaging", an emerging field based on the revelation that a small number of linear projections of a compressible image contain enough information for image reconstruction and processing. The implications of compressive imaging are promising for many applications and enable the design of new kinds of imaging systems and cameras. For more information, a compressive imaging resource page is available on the web at