Application of wavelet denoising to improve compression efficiency while preserving integrity of digital micrographs.

Modern microscopy methods require efficient image compression techniques owing to collection of up to thousands of images per experiment. Current irreversible techniques such as JPEG and JPEG2000 are not optimized to preserve the integrity of the scientific data as required by 21 CFR part 11. Therefore, to construct an irreversible, yet integrity-preserving compression mechanism, we establish a model of noise as a function of signal in our imaging system. The noise is then removed with a wavelet shrinkage algorithm whose parameters are adapted to local image structure. We ascertain the integrity of the denoised images by measuring changes in spatial and intensity distributions of registered light in the biological images and estimating changes of the effective microscope MTF. We demonstrate that the proposed denoising procedure leads to a decrease in image file size when a reversible JPEG2000 coding is used and provides better fidelity than irreversible JPEG and JPEG2000 at the same compression ratio. We also demonstrate that denoising reduces image artefacts when used as a pre-filtering step prior to irreversible image coding.