In recent years algorithms based on wavelet models have become the standard for doing image analysis. Their primary advantage, over Fourier and cosine transform models, is their time-frequency signal decomposition and better energy compaction.

Yet, separable wavelets, used by the majority of today's researchers, do not provide the most useful image decomposition. Separable wavelets do not adapt to the local edge, at least not in the sense that the value of a certain wavelet coefficient has direct information on the local edge direction, which is critical for developing efficient image interpolation, denoising, and compression algorithms.

To address these shortcomings, DMMD is developing new image models that provide true 2-D, locally adaptive sets of basis functions. The basis functions are aligned along the local edge and provide better energy compaction, and local directional information than separable wavelets. The new image models allow for improved algorithms in areas such as image denoising, interpolation, compression and many other applications.

Results
Here is a comparison of DMMD Image Enhancement Algorithms against other currently published methods: