Application of detector precision characteristics for the denoising of biological micrographs in the wavelet domain. (May 2015)

• Record Type: Journal Article
• Authors: Bernas, Tytus; Starosolski, Roman; Wójcicki, Robert
• Is Part Of: Biomedical signal processing and control. Volume 19(2015)
• Page Start: 1
• Page End: 13
• Publication Date: 2015-05
• Subjects: Microscope imaging -- Biomedical imaging -- Acquisition noise modeling -- Wavelet image denoising -- Adaptive soft thresholding; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2015.02.010
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: We propose a new wavelet-domain denoising algorithm for biological micrographs. We compute noise characteristics directly from image acquisition parameters. Algorithm exploits intra- and inter-scale correlation of wavelet coefficients. Algorithm outperforms others: AdaptShrink, BivarShrink, and OraclShrink. Abstract: Typical fluorescence microscopy images contain large amounts of noise, which depends on the signal in a complex manner. This characteristic is substantially different from digital photography or satellite data, for which most of the existing denoising algorithms have been designed. Therefore, an efficient estimation of the noise in fluorescence micrographs and its removal pose a challenge. On the other hand, as shown previously, the use of a calibrated microscopy detector may allow computation of the signal and noise characteristics directly from the image acquisition parameters. Therefore, we propose a denoising algorithm that takes advantage of this information to obtain an estimate of the signal and the corresponding noise in the wavelet domain. This general model was constructed using actual fluorescence micrographs and utilizes intra- and inter-scale correlations of the wavelet coefficients. The signal-to-noise estimate was then applied to perform local soft thresholding in the wavelet domain. The performance of the proposed algorithm was tested using a set of images of several common subcellular structures containing various amounts of signal-dependent and signal-independent noise. The denoising performance of the new algorithm depends on the actual amount of noise and on the type of imaged structures. In every case, we demonstrated that the proposed algorithm is superior to two other locally adaptive denoising algorithms (AdaptShrink and BivarShrink) and to optimal subband adaptive soft thresholding (OraclShrink).