A signal acquisition and recovery framework asserting that sparse or compressible signals can be recovered from far fewer linear measurements than traditional Nyquist sampling dictates, by designing incoherent measurement operators and solving sparse recovery problems (via l1 minimization, greedy pursuit, or combinatorial algorithms) that exploit the signal's sparsity structure.

Principle

Exploit sparsity or compressibility: if the signal has a sparse representation in some basis and the measurement operator satisfies incoherence or Restricted Isometry properties, then stable and robust recovery from undersampled linear measurements is possible by promoting sparsity in the reconstruction.

Demonstration

Acquiring a k-sparse image patch with m<

Misapplication

Designing measurements that are coherent with the sparsity basis (e.g., sampling only the same few basis vectors) or applying sparsity-promoting recovery when the signal is not well approximated as sparse, causing poor reconstructions and misleading confidence in undersampled data.

Consequence

When assumptions hold, compressive sensing reduces sampling and storage requirements and enables new acquisition systems (single-pixel cameras, sub-Nyquist ADC architectures) by shifting complexity into computational recovery, but it requires careful measurement design and computational resources for reconstruction.

Reversal

Classical sampling and reconstruction: sample at or above the Nyquist rate and reconstruct by direct inversion or interpolation methods without leveraging sparsity; this requires many more measurements but avoids the need for sparsity-based recovery algorithms and associated computation.

Boundary

Scope: signals that are exactly sparse or well-approximated by sparse representations in a known basis or dictionary, with measurement ensembles or operators satisfying incoherence or RIP-like conditions; excludes arbitrary signals without compressible structure, measurement processes that cannot be randomized or designed, and scenarios where reconstruction cost or latency is prohibitive.

Semantic Tension

Tension between measurement complexity and reconstruction robustness: more structured, hardware-friendly measurement designs (e.g., partial Fourier, structured random matrices) ease implementation but may require stronger incoherence assumptions or more measurements compared to fully random ensembles that are information-theoretically stronger.

Synthesis

Compressive Sensing is a theory and methodology that trades measurements for computation: by designing measurements incoherent with a sparse representation and using sparsity-promoting recovery algorithms, it enables accurate reconstruction from substantially fewer linear samples than classical sampling would require under suitable structural assumptions.