Enhance Image (AI)
Use the GeoDict-AI module to train a neural network to transform low-quality images into high-quality ones. For example, scans taken from fewer angles or with reduced exposure time can be improved by a trained network. The GeoDict-AI user guide explains how to train a neural network to enhance gray value images.
If a suitable graphics card is detected, the GPU mode is used to run Enhance Image (AI). Otherwise, it runs in CPU mode. If multiple GPUs are available, select the one on which it should run. Image enhancement can also run on multiple GPUs. If you select more GPUs than are licensed, an error message will appear when you click Apply.
Browse for the neural network model trained with GeoDict-AI and best suitable for the loaded image dataset.
The Description lists the current constraints for applying the neural network, as entered in the GeoDict-AI module for training neural networks. For example, it specifies the type of images that the neural network was trained with.
The Batch Size selection is related to the available memory on the graphics card (GPU). Conceptually, GeoDict loads the GPU with portions of work called batches. Currently, this selection must be made manually, and the parameter is set according to the value entered in Batch Size. The larger the Batch Size, the better the performance. Therefore, set the Batch Size as high as possible, depending on your GPU and the neural network settings. If multiple GPUs are selected, the batches are distributed equally among them. If a GPU is not detected, image enhancement runs on the CPU, which takes much longer.
If the Batch Size exceeds the graphics card's capabilities, an error message appears and the image enhancement process is aborted.
Click Apply to start the image enhancement.
|
Know how! You can Save your current settings for this tool as Start-Up Settings . The next time you process an image, these settings will be automatically loaded and filled into the parameter fields. You can also load the Built-In Default Settings available in GeoDict. If you change any settings and want to revert to your saved start-up settings, click the corresponding button to Load the Start-Up Settings . |
Example
A carbonate sample was scanned twice, once with a short exposure time and once with a long exposure time. The sample data was provided by Chandra and Al-Naimi. A neural network was trained using GeoDict-AI to transform the low-quality scan into a high-quality scan. Below is a comparison of applying the NLM Filter and the trained AI model to the low-quality scan. While the NLM filter reduces noise, the contrast is worse compared to the scan with a long exposure time. However, the AI model produces a scan with a comparable contrast to the high-quality scan and has less noise. Once a model is trained with GeoDict-AI, short exposure times are sufficient to obtain high-quality images using the Enhance Image (AI) tool.
