Data compression for diffraction patterns

Abstract

Efficient coding (lossless) and compression (lossy) of diffraction patterns is important in protein crystallography experiments because of storage and transmission limitations. The goal is to reduce the bit-rate significantly while keeping diffraction peak intensity distortion at an acceptable level. This paper presents an overview of coding and compression techniques more or less adapted to such problems. A large part of this study is dedicated to time-frequency-transform based compression algorithms and some of their extensions. Wavelet based software has been developed and tested. Results are compared with the discrete cosine transform (DCT) and other classical algorithms. These tools seem attractive and very promising for analyzing and compressing signals with singularities and transient phenomena such as diffraction peaks. Tests were performed on a standard protein crystallography data set coming from the CCD detector of D2AM beamline at the European Synchrotron Radiation Facility at Grenoble. These were compressed with DCT and wavelet-based algorithms. It appears that alterations of the result of the processing of restored images remain very weak for compression rates up to 10. These preliminary results indicate that the proposed wavelet method is a good standard technique for efficient compression of diffraction patterns.