Hypermap Segmentation

requires amics 5 2 0 or newer amics 5 2 is currently in beta testing the released version of amics is {{version}} and can be found with the most recent beta at the docid\ e1faehd1k20uedqhptzlx page manuals for previous versions of the software can be found at the docid\ d2biwe sybowhhdww0anz page introduction this function allows users to perform post segmentation of the data using spectral distance as the segmentation variable, rather than using a bse watershed function, as in segmentation mapping as this function is designed to segment the data using spectral data, it is best used for sem amics hypermap analysis or m4 amics data though the function will also work for sem amics matrix mapping, it would not be appropriate for sem amics segmentation analysis modes (segmentation mapping, rare phase search, or image mosaic) the purpose of this function is to reduce the post processing requirements of m4 and hypermap data, which tends to include a larger number of mixed spectra, by using the spectral distance to allow for some deviation from the pure material spectra it should be noted, however, that this function will not remove all mixed spectra from the data set theory the hyperseg function works by first removing the background pixels using the values set by the users these pixels will no longer be part of further downstream processing and will be assigned to the special material class background from the remaining pixels, a seed spectrum is selected, and the seed pixel's spectrum and bse/total energy intensity are then compared to those of its near neighbors this process continues until either a change in the bse/total energy intensity or a matching of the spectra exceeds the threshold set by the user in addition, if a minimum/or maximum count is set, these thresholds will also be used to terminate a segment all pixels that match the seed pixel within this threshold will then be merged into a single segment, with a summed spectrum assigned to the segment usage the hypermap segmentation function is accessed from the processing ribbon, and using the hyperseg button running the hyperseg function using this button will bring up the resegmentation options pop up menu this menu will provide the setting for resegmenting the amics data set using spectral distance the settings for segmentation are as follows hyperseg filter option min bse sets the lower bse (total energy intensity) value for valid materials max bse sets the upper bse (total energy intensity) value for valid materials note! these two bse settings will define the bse or total energy intensity value, which will be used to describe the materials in the dataset, and is not the background this allows for a lower and upper background to be determined, as any pixel with a value lower than the minimum value and greater than the maximum value will be defined as background and not segmented hyperseg settings option max bse distance sets the allowable difference of the bse or total energy intensity between two spectra points matching ratio define the allowable spectral difference defined as the same segment set to " 1" to disable setting min total counts sets a minimum total count required per segment set to " 1" to disable setting once the options are set and the resegmentation is executed, amics will pop up a new menu asking how to handle the data (see image below) the options are current will make changes to the opened data set and overwrite the original data another will clone the data set first, and process the cloned data set cancel will terminate the hyperseg function without executing it hyperseg example in the example below, a simple m4 data set is being used, which contains only three materials two minerals and the background resin in this example, the background will be set to anything lower than 50 for the total energy intensity, so the two materials shown are between 50 256 the image above also shows the original segmentation image before executing the post segmentation routine in this image, each pixel is classified as a segment, resulting in a multicolored, pixelated segmentation map in this example, the above settings have been used the settings will cut out all pixels with a bse/total energy intensity value of 0 49, which will be defined as background the maximum difference in bse/total energy intensity value between has been set to 100; therefore, nearest neighbours must have a bse/total energy intensity difference of less than 10 in this example, a matching ratio of 0 1, equivalent to 10 in the amics classification setting and no minimum or maximum total counts have been set the result from the example hyperseg is shown in the above images, which shows that nothing has changed in the bse/total energy intensity images however, the classified map is now showing background (i e , white space), with a significantly reduced unknown grey area compared to the original dataset the significant difference is evident in the updated segmentation map, which now displays areas of similar spectra being combined into segments note the hyperseg routine will only resegment the data set; it will not convert the data from block to granulated sample type this means that the converted data may still only show the frame option, but to maximize the value for hyperseg of a particulated sample, it is advisable to convert to a granular sample type using the particulator function