eSRRF Advanced parameters
These parameters allow the setting of how the GPU is exploited during the reconstruction. Allowing more resources (large amount of memory on GPU device and large analysis block size) will speed up the reconstructions but will require a more powerful GPU and/or slow down other ongoing processes on the machine.
Please note that the plugin has currently been tested on the following configurations:
- MacBook Pro with embedded GPU (Intel(R) Core(TM) i9-9880H CPU @ 2.30GHz, AMD Radeon Pro 5500M Compute Engine)
- Windows 10 machines with Nvidia GTX 1080, 1060 & Nvidia GeForce GTX 1660 Ti with a 516.94 driver, DirectX runtime version 12, and Nvidia CUDA 11.7.101 driver.
- Windows 10 machine with Nvidia GeForce RTX 3080 GPU with 528.49 driver and Nvidia CUDA 12.0.1 driver
- Windows 11 machine with Nvidia RTX 4090 GPU with 531.29 driver, DirectX runtime version 12, and Nvidia CUDA 12.1.68 driver
This option will allow the user to save the reconstructions directly to disk, and therefore saving RAM requirements. This is especially useful for reconstructions of large videos. It currently only works for 2D analyses.
These additional options are only for testing purposes and should be kept as default unless a specific component need to be tested. This is mainly aimed at developers.
- Disabling the intensity weighting allows the observation of the pure raw unweighted radial gradient convergence maps.
- The following option determines the type of interpolation that is performed during the radial gradient convergence calculation: the default and most favourable option is via FHT (Fast Hartley Transform), otherwise it will be done using bicubic interpolation as was done in the original SRRF method. The 3D-eSRRF reconstruction currently uses the bicubic interpolation.
- The macro-pixel patterning sometimes appears and can be estimated and partially removed. This is only the case when using bicubic interpolation and is therefore disabled when using FHT interpolation.
If the user wishes to display additional intermediate data. It is possible to obtain the interpolated gradients. The show image stabilisation plot option shows a scatter plot of the estimated drift measured in the raw data. The time is colour-coded.
