Deconvolution in 3D: an ImageJ plugin

During this master project, an ImageJ plugin permitting to deconvolve 2D and 3D images has been developed. The implemented deconvolution algorithms are classical algorithms such as Landweber or Tikhonov-Miller, and state of the art algorithms such as Richardson-Lucy with total variation regularization. The point spread function (PSF) is given to the plugin as an image, which enables the plugin to address a wide range of deconvolution problems, as for example out-of-focus blur or motion blur.

Furthermore, additional features facilitating the tuning of algorithm-specific parameters have been implemented, as the user has for instance the possibility to deconvolve only a small ROI of an image, or to save automatically the intermediate results of an iterative algorithm. The optional use of the FFTW library gives the possibility of speeding up the deconvolution process considerably.

While the main target group was the biological microscopy community, the plugin provides also a good framework for developing and testing new algorithms, making it a useful tool for the image processing community as well. A view of the plugin's GUI, as well as an example can be found in figures 1 to 4.

Fig.1 Input image-stack taken by a simulated wide-field microscope.

Fig.2 Deconvolved image-stack, after 40 Richardson-Lucy iterations.

Fig.3 GUI of the ImageJ plugin "Deconvolution".

Fig.4 GUI of the ImageJ plugin "PSFGenerator".