Picture coding in video data transmission based on highly compressive vision models

This thesis describes a highly efficient recursive method of image compression based on a pyramid of reduced-bandwidth and reduced-resolution versions of the image. This method employs the Gaussian-Laplacian kemels to convolve with the image. First, pixel-to-pixel correlations are removed by subtracting a low-pass filtered version of the image from the image itself. The result should be a net data compression since the error image has low variance and entropy, and low-pass filtered image represented at reduced sample density. Iteration of the process at appropriately expanded scales generates a pyramid data structure. Further data compression can be achieved by quantizing the error function. Fast algorithm for Gaussian-Laplacian pyramid technique has been implemented in a microcomputer-based system for coding and decoding. The main advantages of this algorithm are:

1. It requires a comparatively small amount of computer memory; this makes it easy to implement in microcomputers.
2. It has less computational steps and fast processing speed.
3. It closely resembles the Gaussian probabiUty distribution. Hence, it performs near-ideal low-pass filtering as well as data compression.

A further advantage of the Gaussian-Laplacian pyramid is that it is well suited for many image processing and analysis tasks as well as for image compression.