Brick splitting

Situations occur where your system's memory is not sufficiently large to allow an image to be deconvolved as a whole (see Not Enough Memory Available). In these cases the Huygens Software splits the image into bricks. The bricks are deconvolved one by one; the results seamlessly fitted together again. The brick splitting routine considers many variables to find the best compromise between available memory and deconvolution requirements. There is a limit for that, and if the bricks become so small that the deconvolution will be a disaster, it won't run.

Even when your system has a large amount of memory, the image is split in situations where Spherical Aberration is present, to adapt the Point Spread Function to the depth inside the image. A different PSF is calculated for each brick at different depths due to the fact that the Mismatch Distorts the PSF.

Brick control and errors

To find out the best number of bricks, let the software run in automatic mode for splitting. It will consider many options and go for the best one.

Mind that if your Microscopic Parameters are not physically realistic this brick routine can also fail. The routine expects that the Point Spread Function will be smaller or at most as large as the original image. That is the regular case. But if the deconvolution algorithm needs to calculate a Theoretical PSF based on your Microscopic Parameters, and these provoke the PSF to be huge, the brick routine may not do a good job and you may encounter a 'Not Enough Memory Available' error. This happens for example if your microscope type is a Nipkow Disk Microscope and you set a Back Projected Pinhole Distance larger than the image size, which is very unrealistic.

Still, if you find that the bricks are not doing a good job, you can send us the execution log with the parameters the software considered. Please refer to Execution Log for more details. Run deconvolution until the very end and send us the report.

New way to handle PSF shape changes due to spherical aberration

The shape of PSFs suffering from spherical aberration due to refractive index mismatch is dependent of their depth position in the specimen. Previously this was solved by splitting the image into 'bricks', each with their own PSF. Though efficient, this could give rise to difficult to avoid gluing artifacts. With a 'sub-brick' (VarPSF) method, the Fast Fourier Transform (FFT) based convolution operations which are the heart of all deconvolution algorithms, are replaced by multiple FFTs, each with different PSFs. This new technique is available for the CMLE, QMLE and the new GMLE deconvolution algorithms. It was introduced in Huygens 4.5 for STED data.