profoundAutoMerge: Segmentation Map Auto Merging

In asgr/ProFound: Photometry Tools

Segmentation Map Auto Merging

Description

Attempts to de-fragment large flocculent galaxies that tend to be broken up, even with reasonable watershed parameters. Ideally the settings should be chosen such that you do not go too far the other way, and over-merge objects again.

Usage

profoundAutoMerge(segim, segstats, spur_lim = 0.004, col_lim = NULL, Ncut = 1)

Arguments

segim	Integer matrix; required, the segmentation map. This should usually be the segim_orig (un-dilated) segmentation map, because that usually behaves better with this function. Dilated segmentation maps can also be proved though.
segstats	Data.frame; the segstats output of ProFound. If you want to use the col_lim option, you need to add a non-standard column representing your magnitude colour of choice. A useful colour is usually optical minus NIR, e.g. (r - Z).
spur_lim	Numeric scalar; the threshold in segstats$cenfrac / (1 + segstats$sep) below which when selecting segments as potential merging candidates. The default works well for optical and NIR images, but will require some tuning if the seeing is very far from 1 arc-second.
col_lim	Numeric vector; if provided this is the lower and upper colour limit to use when selecting segments as potential merging candidates. If this is used then the 'col' column must be supplied in segim. If the colour provided is optical - NIR then a range of between 0 and 1 is usually sensible for picking out true galaxy segments.
Ncut	Integer scalar; at least this many segments need to be touching to be provided in the final output. Ncut or 1 or 2 usually makes most sense.

Details

This function should not be applied blindly and requires some though and tuning on images obtained from a new facility. The default works well for good seeing optical imaging from e.g. KiDS.

Value

The output is exactly the same as the groupsegID from profoundSegimGroup. For reference see below.

groupsegID

A data.frame of lists giving the segIDs of segments in each group.

The data.frame returned by groupsegID is a slightly unusal structure to see in R, but it allows for a compact manner of storing uneven vectors of grouped segments. E.g. you might have a massive group containing 30 other segments and many groups containing a single segment. Padding a normal matrix out to accommodate the larger figure would be quite inefficient. It contains the following:

mergeID	Merging group ID. This is different to the usual groupID since it is possible to have islands that have been identified to be re-grouped that lie within superset groups.
segID	An embedded list of segmentation IDs for segments in the merging group. I.e. each list element of segID is a vector (see Examples for clarity).
Nemrge	The total number of segments that are in the merging group.
Npix	The total number of pixels that are in the merging group.
groupID	Group ID of the superset group based on the provided segim.

Author(s)

Aaron Robotham

See Also

profoundSegimGroup

Examples

## Not run: 
image = Rfits_read_image(system.file("extdata", 'VIKING/mystery_VIKING_Z.fits',
  package="ProFound"))

profound = profoundProFound(image, magzero=30)

#Below is just to see how it works, this is clearly over-merging.

groupsegID = profoundAutoMerge(segim=profound$segim_orig, segstats=profound$segstats,
  spur_lim=3e-2)

segimfix = profoundSegimKeep(segim=profound$segim, segID_merge=groupsegID$segID)

## End(Not run)

asgr/ProFound documentation built on Feb. 10, 2024, 9:04 p.m.