This interface allows you to setup and start recipe step (process) optimization (i.e., auto-setting determination). After selecting optimization parameters, you will be asked to enter a range for some of the process’ parameters. This will start a loop which will execute the process at each parameter combination within the specified range. The parameter combination which yields the best quality metric will be set as the new parameter combination for the process. This calculation will be re-performed for any image the Recipe is run on.
The optimization interface is available by clicking the the “Optimize” button in the top right corner of the Recipe panel. The following steps are optimizable:
- Rotate Image
- Adjust Contrast
- Basic Threshold
- Range Threshold
- Adaptive Threshold
- E-M Threshold
- Local Threshold
- Region Grow
- Fast Marching Method
- Find Edges
- Dilate Uniform
- Dilate Smart
- Dilate Retain
- Erode Uniform
- Erode Smart
- Separate Features
- Smooth Features
- Reject Features
1. Comparison Image
Image to compare processed image to during optimization
2. Smart Reference
Choose whether to apply high-pass FFT filtering to comparison image. When “On” Smart Reference with make feature edges more important, and often improves the results of your optimization.
3. View Images
Choose whether to display optimization at each step
4. Save Images
Choose whether to save each frame of the optimization movie
5. Save Histograms
Choose whether to view/save the histograms of the processed image at each step
6. Area of Interest
Choose whether to use a sub-region or the entire image for optimization
Choose metric to use for comparing processed image to the comparison image
- Mutual Information: A measure of image similarity that describes how well the processed image describes the comparison image [1,2]. Takes into account the uncertainty of the comparison image. The optimization will choose the parameter combination which yields a maximum mutual information.
- Intra-class Variance: A more basic measure of image similarity that describes how well the processed image describes the comparison image. Does not take into account the uncertainty of the comparison image. Better suited for simpler segmentation problems with clear contrast between features and background. The optimization will choose the parameter combination which yields a minimum intra-class variance.
- Number of Features: Optimization will monitor the number of features that result over the explored parameter range. The optimization will choose the parameter combination which yields the flattest derivative (plateau) in the number of features vs. parameter value curve. This can be useful for optimizing functions like Separate Features, where you may want to choose the weakness value which produces the the most stable number of separate features.
Choose whether to use the images, or their two-point correlation functions for comparison
9. Parameter Value (min)
Minimum value of the parameter range to explore
10. Parameter Value (max)
Maximum value of the parameter range to explore
11. Parameter Value (inc)
Increment (step size) to take between the minimum and maximum parameter values
 Rahunathan, Smriti, D. Stredney, P. Schmalbrock, and B.D. Clymer. Image Registration Using Rigid Registration and Maximization of Mutual Information. Poster presented at: MMVR13. The 13th Annual Medicine Meets Virtual Reality Conference; 2005 January 26–29; Long Beach, CA.
 D. Mattes, D.R. Haynor, H. Vesselle, T. Lewellen, and W. Eubank. “Non-rigid multimodality image registration.” (Proceedings paper).Medical Imaging 2001: Image Processing. SPIE Publications, 3 July 2001. pp. 1609–1620.
Need more help with this?
Chat with an expert now ››