摘要
This paper presents a scheme for improving encoding time for fractal image compression. The approachcombines feature extraction with domain classification using a selforganizing neural network. Feature extractionreduces the dimensionalics of the problem and enables the neural network to be trained on an image separate fromthe test image. The seaorganizing network introduces a neighborhood topology for classytcation, and alsoeliminates the need to specify a prior set of appropriate image classes. The network organizes itself according to thedistribution of the image features observed during the training. The paper presents results showing that thisclassification approach can reduce encoding time by two orders of magnitude while maintaining comparableaccuracy and compression performance.
This paper presents a scheme for improving encoding time for fractal image compression. The approachcombines feature extraction with domain classification using a selforganizing neural network. Feature extractionreduces the dimensionalics of the problem and enables the neural network to be trained on an image separate fromthe test image. The seaorganizing network introduces a neighborhood topology for classytcation, and alsoeliminates the need to specify a prior set of appropriate image classes. The network organizes itself according to thedistribution of the image features observed during the training. The paper presents results showing that thisclassification approach can reduce encoding time by two orders of magnitude while maintaining comparableaccuracy and compression performance.