Gallery: Binning

This example produces the same output as the example above, but rather than specifying the bin size we specify the number of bins.

In this example, the individual columns X and Y ranges have been replaced with the simplified syntax [bin sky=2]. sky is the name of the vector column, composed of the X and Y columns.

By default the image shrinks to a bounding-box around the filter. Values outside the region are set to 0. This behavior can be changed using DM options.

The image on the Left shows the grid that dmnautilus computed, the 6934_mask.fits output file. Each of the cells in the grid has an upper limit SNR=20 (400 counts) -- single pixel cells that could not be subdivided further can exceed this limit.

The image on the Right shows the data with that binning applied. The boundary of the cluster emission is more easily identified.

The image on the Left shows the grid that dmradar computed, the 6934_rmask.fits output file. Each of the pie wedged in the grid has an upper limit SNR=20 (400 counts).

The image on the Right shows the data with that binning applied. The boundary of the cluster emission is more easily identified.

In this example we have a input map (Left), 6934_s3.contour.map . The pixel value indicates to which group the pixel location belongs.

Right: The output image, 6934_cbin.fits with a contour mapping binning applied. The sum of the pixel values in each group from the input counts image have been normalized by the size of the group and that average value replicated in each pixel in the group in the output image.

In this example we take the event file for CasA and apply the spatial binning and energy filtering we saw in the earlier examples.

We then run dmnautilus on the broad band, ie combine hard+soft bands, to create a grid where we have a combined SNR limit of 20 (ie a limit of 400 counts).

We then use this map to create an image of the soft band and hard band counts.

And the finally we combine those two images using the equation above to compute the hardness ratio map.

Since we have selected the bins to include roughly 400 counts each, the statistical error in this map is low.

We can clearly see from this map the hard emission in the Western part of the remnant. (North is towards the top of the image).

This thread starts by using dmimgpick to add a new column to the event file. The new column is called cas_a_broad_map. The tool looks up the sky position for each event and then uses the closest value in the map file to populate the new column.

Next we use dmcopy to create an image but not in sky coordinates. In this example the X-axis is the value of the cas_a_broad_map column; ie the group ID. The Y-axis is binned on pi, which is the observed energy channel.

The results: We now have an image where each column is the spectrum for different parts of the sky. The image above shows just a small segment of the total image zoomed into a region where we can clearly see the change in the spectral signature.

Note: This type of spectrum is not immediately useful with CIAO, as we do not have the appropriate response files (ARF and RMF), nor do we have the information we need to account for background (including area scaling information).

In this example we created a hexagonal grid where each hexagon is has a side length equal to 10 pixels. The input image was filtered with the field-of-view file so that the tool knows where the edge of the detector is and leaves those pixels ungrouped (map value equal to 0). The map file is shown in the left-hand side of the image. For the color map we have chosen a random set of colors to help highlight the different map ID values.

The output map output is then run through the map2reg script to create a region file from the map grid; one region is created for each map ID value found in the input map file. Note: since the map ID values are not necessarily continous nor do they necessarily start at 1, the row number (not component value) in the region file may not correspond to the map ID value in the input file. This file can be used to easily display the map grid as is shown in the right-hand side image.

As is shown in other Gallery examples, the dmmaskbin tool is then used to apply the map file (unfortunately the terms "mask" and "map" are sometimes used interchangeably) to the original counts image which is what is display in the right-hand side image.

In the left hand frame we see the map created by vtbin. As with the hexgrid example, the pixel values represent the map ID for each region; the color map is a set of random colors to show the different regions. The right hand frame shows the output image after the map file has been applied to the original input image using the dmmaskbin tool with the region file created by map2reg overlaid.

The choice of the initial central points is the key to this algorithm. The default is to use the local maxima however users can supply their own set of points which is used by the centroid_map script.

This is similar to the vtbin example showing the map file on the left and the binned image on the right. The polygons are now more uniform in size and spacing. For very bright sources, the polgyons can become very small; users can use the scale parameter to select different pixel scaling when computing the centroid which will lessen the impact of bright pixels.

In this example of the star forming region in Orion, there are many individual point sources and with diffuse emission in the center. The algorithm is able to separate many of the individual stars based on the gradient searching algorithm. The left image shows the map file with a random color map applied. The right image shows the output by binning the original image with the map file.

The amount of smoothing can have significant impact on the algorithm. Using an adaptive smoothing tool like dmimgadapt or csmooth may help to provide even better source separation. Subtracting the background may also help to resolve all the point sources embedded in the extended emission.

The map file (left image) created from the mkregmap script shown with random colors. The right image shows the map file applied to the original image. Since the region files are actually input to the script there is not need to use map2reg to create them; we can just directly overlay the contours on the binned image.

Question: How did we select the number "18" in the igrid(1:18:1) syntax? Answer:

unix% dmlist ell.fits counts
18