Figure 589. 3D volume data before and after deconvolution
There are different ways to sort deconvolution methods, either by the image data to be processed (2D/3D) or by the method of PSF determination (blind, Richardson-Lucy), or by expected quality of the result (fast). There are several commands available within the Deconvolution pull-down menu. Read the following overview and decide which command best suits your requirements:
Deconvolution Commands
(requires: 3D Deconvolution) This command is designed to work strictly with volume data-sets i.e. ND2 files containing Z-stack.
(requires: 2D Deconvolution) Standard 2D single-frame images or ND2-files (Timelapse, multipoint, Z-stack...) can be processed by this command.
Note
Frame-by-frame deconvolution is performed in this case. Z-stack ND2 file can be processed by this deconvolution method as well, but the result will not be as good as you get from the 3D method.
Deconvolution of the live image may be performed by this command. A control panel appears where its parameters can be set and the processing on live image can be turned on/off.
Deconvolution methods
It is possible to decide between the following deconvolution methods. Each method uses a different algorithm to achieve the same goal (see Algorithms).
It is stable method which produces good results.
It is faster than blind deconvolution.
It reduces image noise efficiently.
It produces decent results even if you do not import PSF.
The results may have lower contrast compared to the Blind deconvolution.
Some image details may remain blurry.
It is the fastest method (compared with Blind and Richardson-Lucy methods).
It is stable (little less than the Richardson-Lucy method).
It produces excellent results (not as bright as Blind in most cases, but with significantly less noise, which makes sample structures more clear).
Needs approximately the same memory as Richardson-Lucy method.
Rarely diverges.
It produces top quality results.
It makes very precise estimation of the PSF shape.
It enhances even the smallest image details.
The strength of this method may amplify image noise.
In some cases, the method might diverge and give unsatisfactory results.
It is super fast
Despite its swiftness, it can produce good results.
The results are not as good as from the other methods.
It can add artifacts to the image.
It can handle frame-rates up to 10 fps.
It can be applied to a restricted area of the image (ROI).
Speed is the priority at the expense of image quality.
This method automatically deconvolves the opened image based on its metadata. If proper metadata are not available, deconvolution parameters can be specified before running the deconvolution process.
Using Richardson-Lucy deconvolution assumes we know the PSF exactly, so the system does not have to estimate it (see Determining the Point Spread Function (PSF)). This type of deconvolution usually leads to reliable results.
Landweber deconvolution is a current state-of-the-art method based on wavelet deconvolution. It can deal with noise and deliver crystal clear results.
This deconvolution method estimates the shape of the PSF from the parameters specified within the deconvolution dialog window. However, if the known PSF is used, it further improves the final result. Concerning RAM usage, it is the most demanding of all the methods.
This method is optimized for speed.
This method uses deconvolution algorithm similar to fast deconvolution but can be used on Live image (stream).
Figure 590. Comparison of deconvolution methods, from left: original image, fast, Richardson-Lucy, blind
Figure 591. Comparison of deconvolution methods, from left: original image, fast, Richardson-Lucy, blind
Figure 592. Comparison of deconvolution methods, from left: original image, fast, Richardson-Lucy, blind