Abstract

A method is described for encoding a first image to produce a second, lower resolution image. For each superpixel group in the first image, a prediction class is formed, based on the values of the pixels in two adjacent superpixels. This prediction class is used to access a table, to obtain a prediction for the corresponding pixel in the second image. If the prediction matches the actual pixel pattern of the superpixel, the corresponding pixel in the second image is set to the predicted colour. If the prediction does not match, then the corresponding pixel in the second image is set to the inverse of the predicted colour, and the actual pixel pattern is stored in a supplementary file. If no prediction is provided, the actual pixel pattern is stored, and the pixel in the second image is set to the colour of the majority of the pixels in the superpixel. The result is a compressed, low-resolution image, along with a supplementary file containing the information necessary to reconstruct the high-resolution image from the low-resolution image. The method may be applied repeatedly so as to form a series of images of progressively lower resolution.