Abstract
A new method for image data compression and reconstruction is described. In this method two-dimensional interpolative linear estimation is employed to obtain a large data compression ratio. The gray level of each pixel is estimated by taking a linear weighted summation of gray levels of its neighbours on all sides. The weight coefficients are determined so that the mean square estimation error is minimized.
The optimal estimation errors are evaluated pixel by pixel, and the obtained error image is added to the original image multiplied by a small positive weight. Then, this compressed data are transmitted together with the estimator weight coefficients.
To reconstruct the original image from the compressed data, an iterative algorithm based on the conjugate gradient method is used. It is shown that the addition of original image information to the estimation error data reduces drastically the iteration number required in the reconstruction process at a small increase of the data amount to be transmitted.
