Apply online to license this technology
Even with the decreased cost and increased availability of both bandwidth and storage space, transferring and saving images and video is still time-consuming. Data compression methods such as JPEG, JPEG 2000, or MPEG-4, and others are bandwidth and space hogs, and as we flow more robust data over networks, we need better compression techniques.
The Shannon sampling theorem dictates that any signal needs to be sampled at the Nyquist rate (2x the maximum cut off frequency at which the bandwidth of the signal tends towards zero). The theorem has guided data compression approaches for years, but it is theoretically valid only for a stationary signal whose second-order statistics do not vary with time. In real life most signals are non-stationary, and the spectral content, the center frequency, and the bandwidth of the local signal vary with time. Given this, researchers have attempted several approaches to sub-sample the signal below the Nyquist rate exploiting the property of time-varying spectral characteristics of the signal. Although these attempts have resulted in some improvement in reduced sampling for a class of signals, they have not produced a lossless compression scheme or even near-lossless compression with a significant reduction in sampling rates.
Naval scientists have overcome these obstacles with a way to non-uniformly sample signals below the Nyquist rate without many of the associated aliasing artifacts from previous approaches as the signal can be uniquely reconstructed with negligible, (below noise level), errors using linear interpolation. This compression method can be used in combination with other compression methods.
- Non-uniform sampling of a signal may result in a reduced number of samples much lower than permitted by the Nyquist criterion but that still uniquely represent the signal
- Non-uniform sampling does not cause the traditional folding of the spectrum associated with sub-sampling and aliasing
- With linear interpolation the non-uniform sampling results in an absolute mean percentage error of 0.4%, while the traditional uniform sub-sampling technique results in an error of 19.8% and with cubic interpolation these errors can be further reduced
- This near lossless non-uniform sampling compression method alone results on the average a 50% reduction in storage and coupled with other lossless compression methods can result in 70% reduction in storage or bandwidth or an equivalent compression factor of 4
- Can be incorporated into graphics and video processing chips for improved real time 3D high definition applications
- US patent 8,526,746 available for license