Performance analysis of a new semiorthogonal spline wavelet compression algorithm for tonal medical images.

Lossy image compression is thought to be a necessity as radiology moves toward a filmless environment. Compression algorithms based on the discrete cosine transform (DCT) are limited due to the infinite support of the cosine basis function. Wavelets, basis functions that have compact or nearly compact support, are mathematically better suited for decorrelating medical image data. A lossy compression algorithm based on semiorthogonal cubic spline wavelets has been implemented and tested on six different image modalities (magnetic resonance, x-ray computed tomography, single photon emission tomography, digital fluoroscopy, computed radiography, and ultrasound). The fidelity of the reconstructed wavelet images was compared to images compressed with a DCT algorithm for compression ratios of up to 40:1. The wavelet algorithm was found to have generally lower average error metrics and higher peak-signal-to-noise ratios than the DCT algorithm.

Introduction to wavelet-based compression of medical images.

Medical image compression can significantly enhance the performance of picture archiving and communication systems and may be considered an enabling technology for telemedicine. The wavelet transform is a powerful mathematical tool with many unique qualities that are useful for image compression and processing applications. Although wavelet concepts can be traced back to 1910, the mathematics of wavelets have only recently been formalized. By exploiting spatial and spectral information redundancy in images, wavelet-based methods offer significantly better results for compressing medical images than do compression algorithms based on Fourier methods, such as the discrete cosine transform used by the Joint Photographic Experts Group. Furthermore, wavelet-based compression does not suffer from blocking artifacts, and the restored image quality is generally superior at higher compression rates.