While ImportGeo-Vol has many effective image filters, there are cases, where a neural network performs better. Use GeoDict-AI to train a neural network to enhance grayscale images, for example for CT-scans taken from only a few angles or with reduced exposure time. The trained neural network can either be used with Enhance Grayscale Image or in the ImportGeo-Vol Image Processing dialog.

To train the neural network, you need low-quality / high-quality image pairs.

These must fulfill the following requirements:

• they must be registered, i.e. capture the same region,
• currently, they must have the same resolution and therefore, the same number of voxels in X-, Y- and Z-direction,
• the low-quality scan should be saved as input.grw and the high-quality scan should be named output.grw.

The neural network will then learn to transform low-quality scans to high-quality scans such that ideally you will be able to obtain comparable high-quality results by applying the trained network to future low-quality scans. For this, the files must be organized in the folder structure described here.