High-resolution M-channel, 2-dimension lattice linear prediction algorithm

We present a new high-resolution M-channel, 2-dimensional lattice linear prediction parameter estimation method. The Yule-Walker like equations are formulated, consisting of a nested block matrix structure. The linear prediction parameters are estimated using a Levinson type recursion based on lattice (reflection coefficient) technique. The linear prediction parameters are then used in a companion M-channel, 2-dimensional autoregressive spectral estimator. The output of this technique is sharpened auto-spectral image from the two channels, and super resolved cross-spectral images that preserves what is in common between the two channels. This technique is well suited for Ultrawide-band processing, high-resolution multichannel data fusion and clutter suppression. In the simulation section, we compare two channels with 4 point like sources each, 2 or which are common between the two 2-D channels, and show that sharpened results of the 2 common sources in the combined image.