Image processing

Simple spatial filters can be used to suppress noise in raw image data (i.e. by averaging intensities). The best choice of filter depends on the nature of the noise, but Gaussian filtering works well for Poisson noise (i.e. commonly observed photon-counting shot noise); whereas a median filter is ideal for salt-and-pepper noise. A larger filter radius leads to stronger noise suppression but more blurring. The URL above describes the simple 2D spatial filters available in ImageJ, but similar filters are available in most software. For 3D data, 3D versions of these filters work best (since there are more pixels to average within the same radius).

This macro implements a filter that is meant to attenuate close to parallel intensity stripes in an image, such as often happening in light sheet microscopy. The results are usually decent even when the stripes show a large angular spread due to light sheet refraction at the sample surface. The filter can process a 3D stack but the processing is performed slice by slice.

Example image is available in the documentation link. 

A commercial image analysis software. It's interface allows to easily perform measurements and image analysis. Your actions can be recorded and a macro (in a basic script language) can then be created. Almost no knowledge in programming is needed. You can also use python. A SDK is also available to develop stand alone applications in c++. Additional modules allow to use specific operations (3D operators... Examples of available categories of operators : filtering, edge detection, mathematical morphology, segmentation, Frequency operations, mathematical/logical operations, measurements...

This library gives the numpy-based infrastructure functions for image processing with a focus on bioimage informatics. It provides image filtering and morphological processing as well as feature computation (both image-level features such as Haralick texture features and SURF local features). These can be used with other Python-based libraries for machine learning to build a complete analysis pipeline.

Mahotas is appropriate for users comfortable with programming or builders of end-user tools.

==== Strengths

The major strengths are in speed and quality of documentation. Almost all of the functionality is implemented in for multiple dimensions. It can be used with other Python packages which provide additional functionality.

Mahotas and all packages on which it relies are open-source.