Introduction to Deconvolution

Deconvolution is an image-processing method for suppressing image blur caused by light/optical microscope systems. Deconvolution can be used to create sharper images with better resolution.

Let's answer couple of basic questions before we start using the Deconvolution module:

Q:

What is deconvolution for?

A:

Every microscope optical path (or optics in general) makes the resulting images blurry. Deconvolution is for de-blurring images captured by a digital camera. With NIS-Elements it can be applied to 2D images, 3D volume data, or Live image.

Q:

Which deconvolution method is best for my image sample?

A:

Each method of the deconvolution menu is optimized for different purpose. To choose the right method you have to know especially:

  • Do I need to process live image on the fly or a captured image file?

  • Does the image contain 3D volume data (Z stack) or is it a standard 2D image?

  • Do I need a high-quality result?

  • Do I know exact parameters of the optical path - do I know the point spread function (PSF)?

See Choosing the Deconvolution Method.

Q:

What is PSF?

A:

Every optical path can be described by a so called point spread function (PSF) which specifies how a single point source will look like when captured by a camera. The PSF for each channel is usually represented by a grayscale image. A 2D PSF can look like this:

See Determining the Point Spread Function (PSF).

Q:

Can I run the deconvolution module on my PC?

A:

It is important to realize that deconvolution methods demand appropriate computer performance. It is required to have at least 4GB of RAM, the faster CPU, the better. The need of RAM depends on the size of the processed files, 8GB is a standard, 16GB or more - highly recommended. Crucial acceleration of the deconvolution process can be achieved using a high performance GPU. Multiple GPU graphics cards are supported however it is recommended to use an even number of cards all of the same type.

Note

Deconvolution automatically calculates how much memory is needed for the deconvolution and what can be stored in the memory and what cannot, depending on free RAM memory and the sample dimensions. This optimization makes deconvolution much faster, because the deconvolution can avoid unnecessary accessing HDD.

Note

Implemented deconvolution methods are not suitable for confocal line scan documents.