Cells and tissues are three-dimensional structures, so the observed image at the focal plane contains information from the plane on which the microscope is focused as well as "out-of-focus" contributions from other parts of the specimen above and below the plane of focus. Due to these "out-of-focus" contributions, interpretation of the 3D structure of the biological specimen is hampered in optical imaging. The image at a given focal plane is a poor representation of a true section through the thick specimen. To remedy these problems and produce more reliable 3D data, two methods are widely employed: confocal microscopy and digital deconvolution.

A confocal microscope uses the pinhole aperture to restrict the "out-of-focus" flare reaching a single detector, the photomultiplier tube (PMT). Because of the pinhole aperture, all the fluorescence signal collected by the objective lens is not used. In order to increase the signal, increased excitation laser intensity is used, but this will often introduce photobleaching and photodamage. Another disadvantage of confocal microscopy is its relatively high cost.

Unlike the confocal configuration described above, digital deconvolution microscopy uses the entire fluorescence signal collected by the objective lens without using pinhole to deliver the emitted light to 2D high-sensitivity CCD cameras. The "out-of-focus" flare introduced into the imaging at different optical sections with its subsequent image degradation is reversed by computer deconvolution through the use of a pointspread function (PSF) of the imaging system! By modeling the microscope optics as a linear and shift-invariant system, the PSF can be used to describe the transformation of any image by the microscope. A typical fluorescence microscope image can be: 

[measured image] = [PSF] * [desired image]
where '*' symbol represents the mathematical operation of convolution. The deconvolution (*-1), the mathematical inverse of convolution can be represented as: 

[desired image] = [measured image] (*-1) [PSF] The goal of deconvolution is to solve the equation for the desired image.

The Inverse Filter is a one-step non-iterative approach based upon inverse filtering theory. The deterministic blurring as a convolution of the image with the point spread function can be modeled. In the frequency domain a convolution transforms into a multiplication of the Fourier transform of the sample with the optical transfer function. The optical transfer function (OTF) is the Fourier transform of the point spread function. The inverse filter then accomplishes image restoration by dividing the Fourier transform of the image by the OTF. 

The Nearest Neighbor algorithm works by deconvolving one image slice at a time. It uses information from image slices that reside above and below the image slice that is being processed. The precise increment slice position from the measured slice is user selectable. If the slices are chosen judiciously, this approximation will produce results very close to the inverse filter method but takes much less time. 

The No Neighbor method uses the information from each single slice to construct a 2D PSF. This is the fastest but may not be as representative of the sample as the other methods. 

Inverse filter, nearest neighbor, and no neighbor algorithm functions are all included in SharpStack.

• Image-Pro Plus version 4.5 or higher.
• Pentium III CPU, running at 450 Mhz or higher
• Microsoft Windows™ 98/ME/NT/2K
• 256 MB of RAM (512 MB recommended)
• 3 GB disk drive with available disk space for installation with additional space to accommodate four times the image size with four bytes/pixel
• Color monitor displaying 16-bit high color (24 or 32- bit color preferred)