Mask Generators panel
This panel hosts several ways of automatic mask generation. Specific areas of interest from the generated mask may further be selected for segmentation. See more about masks.
There are several common fields that do not depend on type of the selected mask generator.
- 1. The Filter type combo box, allows to select one of the possible mask generators
- 2. The Mode radio buttons:
- current, generate mask for the currently shown slice;
- 2D all, generate mask for the whole dataset using the 2D mode, i.e. slice by slice;
- 3D, generate mask for the whole dataset using the 3D mode;
the whole dataset in the 2D mode, i.e. slice by slice; 3D use the 3D mode for the mask generation
3. Do it button:
- -left mouse click, starts the selected generator. The existing mask will be deleted.
- -right mouse click + Do new mask, starts the selected generator. The existing mask will be deleted.
- -right mouse click + Generate new mask and add it to the existing mask, the generated mask will be added to the existing mask. This option may be used for multi-dimensional filtering: 1. run Generator for XY; 2. Change dimension by pressing |'XZ'| or |'YZ'| button in the Toolbar; 3. Run Generator again with the |Generate new mask and add it to the existing mask| option.
Hessian based Frangi Vesselness filter, written by Marc Schrijver and Dirk-Jan Kroon. This function uses the eigenvectors of the Hessian to compute the likeliness of an image region to contain vessels or other image ridges, according to the method described by Frangi 1998, 2001.
Note, to work properly this function should be compiled. See details in System Requirements
- Range, the range of sigmas used, default [1-6]
- Ratio, step size between sigmas, default 
- beta1, the Frangi correction constant, default [0.9]
- beta2, the Frangi correction constant, default 
- beta3, the Frangi vesselness constant which gives the threshold between eigenvalues of noise and vessel structure. A thumb rule is dividing the the greyvalues of the vessels by 4 till 6, default ;
- B/W threshold, defines thresholding parameter for generation of the Mask layer. When set to 0 results in the filtered instead of binary image.
- Object size limit, after the run of the Frangi filter removes all 2D objects that are smaller than this value.
- Black on white checkbox, if checked, detects black ridges on white background.
Set of Matlab based morphological filters.
- Extended-maxima transform - based on imextendedmax function of Matlab. Computes the extended-maxima transform, which is the regional maxima of the H-maxima transform. Regional maxima are connected components of pixels with a constant intensity value, and whose external boundary pixels all have a lower value.
- Extended-minima transform - based on imextendedmin function of Matlab. Computes the extended-minima transform, which is the regional minima of the H-minima transform. Regional minima are connected components of pixels with a constant intensity value, and whose external boundary pixels all have a higher value.
- Regional maxima - based on imregionalmax function of Matlab. Returns the binary mask that identifies the locations of the regional maxima in the image. In mask, pixels that are set to 1 identify regional maxima; all other pixels are set to 0. Regional maxima are connected components of pixels with a constant intensity value, and whose external boundary pixels all have a lower value.
- Regional minima - based on imregionalmin function of Matlab. The output binary mask has value 1 corresponding to the pixels of the image that belong to regional minima and 0 otherwise. Regional minima are connected components of pixels with a constant intensity value, and whose external boundary pixels all have a higher value.
Generate mask based on morphological image opening and black-and-white thresholding. The function first performs morphological bottom-hat (Black on white is checked, 5.) or top-hat (Black on white is unchecked, 5.) filtering of the image. The top-hat filtering computes the morphological opening of the image (using imopen) and then subtracts the result from the original image. The result is then black and white thresholded with parameter in the B/W threshold edit box (3.).
- Strel size, defines size of the structural element (disk type) for imtophat and imbothat filtering.
- Fill checkbox, check it to fill holes in the resulted Mask image.
- B/W threshold, specifies parameter for the black and white thresholding.
- Size limit, limits the size of generated 2D objects so that objects smaller than this value are removed from the Mask during the filter run.
- Black on white checkbox, when checked, the filter will use morphological bottom-hat filtering (imbothat). When unchecked - morphological top-hat filtering (imtophat).
Different approaches for black-and-white thresholding. The data should be as bright object on dark background. The thresholding is applied to ROI regions or to the whole image. The resulting mask may be further filtered with tools in the Mask menu.
- For bright object extraction, use the 1. editbox to specify a threshold value. The objects that are brighter than the specified value will be used to generate the Mask layer. The auto checkbox 3. should be unchecked!
- For dark object extraction, use the 2. editbox to specify a threshold value. The objects that are dimmer than the specified value will be used to generate the Mask layer. The auto checkbox 3. should be unchecked!
- Auto checkbox, enables automatic image thresholding using Otsu's method. Note! The auto mode is affected by the values in the 1. and 2. editboxes. If the threshold coefficient is smaller than the parameter in the 1. editbox, the specified 1. parameters will be used instead of automatic coefficient.
- Auto grid checkbox, another approach to the automatic thresholding. In this method the image/ROI region is divided into a separate blocks. The program calculates threshold values for each block and use these values to prepare the image for final thresholding. This method works for cases when the image intensity is constant in each block in Z-direction. The grid will be displayed after the run as the Selection layer and can be removed with Ctrl+C / C shortcut.
- Procedure description:
- 1. Select the |Auto grid| checkbox (*4.*) to enable grid mode.
- 2. Enter size of a single grid box.
- 3. Specify grid threshold coefficient (*5.*). Leave |0| for the automatic mode during the first run.
- 4. Press the *Do it* button to begin the thresholding procedure.
- 5. For each block the proper automatic threshold value is estimated with Otsu's method (this threshold may be altered with values in the 1. and 2. editboxes).
- 6. Find minimal threshold value for the set of blocks or use provided in the 5. editbox coefficient. This value will be used for final image thresholding (thresholdfinal).
- 7. The original image in each box is magnified with the values that
are defined as
Iblock = Iblock x (1 - thresholdblock - thresholdfinal)
- Finally the magnified image is thresholded with the thresholdfinal coefficient.
It is recommended to run grid mode with 0 coefficient in 5.. The image will be somehow thresholded and the used coefficient will be shown in the Matlab command window. If result is not good enough, a new coefficient that is differ from the automatically determined value should be placed into the 5. editbox.
6. Manual grid checkbox, in this mode the coeficients for each box may be manually tuned.
- 1. Run first Auto grid mode. After the run 'thrMatrix' variable with coefficients for each block will be created in the main Matlab workspace
- 2. Select the Manual grid checkbox. Enter the grid size and press the Do it button.
- 3. A new window for manual black and white thresholding will appear. Press the Import button and import thrMatrix variable.
- 4. Press the Start button to theshold the image.
- 5. Adjust the threshold coefficients in the blocks.
- 6. When result seems to be fine select the All check box and press the Start button to make the final thresholding for the whole stack.
The coefficients may be saved and loaded for future use.
Code is optimized for Parallel processing. The grid run mode gives x5-6 times better performance with 8 cores. Parallel processing should be enabled by pressing the Turn on parallel processing button in the im_browser toolbar.