Image De-Quantizing via Enforcing Sparseness in Overcomplete Representations

摘要: We describe a method for removing quantization artifacts (de-quantizing) in the image domain, by enforcing high degree of sparseness its representation with an overcomplete oriented pyramid. For this purpose we devise linear operator that returns minimum L2-norm preserving set significant coefficients, and estimate original minimizing cardinality subset, always ensuring result is compatible quantized observation. implement solution alternated projections onto convex sets, test it through simulations standard images. Results are highly satisfactory terms performance, robustness efficiency.