craverage -- detect CRs and objects using average filter
Craverage detects cosmic rays and objects using a moving block average filter with the central pixel plus some number of additional high pixels excluded and a median of an annulus around the block average box. It avoids identification of the cores of objects as cosmic rays by excluding pixels within the detected objects as cosmic ray candidates.
craverage input output
- List of input images in which to detect cosmic rays and objects.
- List of output images in which cosmic rays are replaced by the block average value excluding the cosmic ray. If no output image name is given then no output image will be created.
- crmask = ""
- List of input and output cosmic ray and object masks. If the mask exists then the mask values are used to exclude data pixels from the calculations and zero mask values are candidates for cosmic rays or objects. Detected cosmic rays and objects are identified in the mask with values given by the crval and objval parameters. If no output cosmic ray mask is given then no mask will be created.
- average = ""
- List of output block average filtered images. If no image name is given then no image will be created.
- sigma = ""
- List of output sigma images. If no image name is given then no image will be created.
- navg = (minimum of 3)
- Square block average filter size given as the number of pixels along an edge. The value will be rounded up to an odd value to be symmetrical around the center pixel excluded from the average.
- nrej = (minimum of 0)
- Number of additional highest pixels to exclude, in addition to the central pixel, in the block average. The value should be small but it is needed to deal with cosmic rays that are bigger than a single pixel.
- nbkg = (minimum of 1)
- Background annulus width around the box average filter in pixels. The median of the pixels in this annulus is used to estimate the background.
- nsig = (minimum of 10)
- Square box size for empirical sigma estimates given as the number of pixels along an edge. The sigma is estimated using percentile points of the pixels in the box. The size of the box should contain of order 100 pixels or more.
- var= 0., var1 = 0., var2 = 0.
- Variance coefficients for the variance model. The variance model is
variance = var0 + var1 * data + var2 * data^2
where data is the maximum of zero and the average filtered pixel value and the variance is in data numbers. All the coefficients must be positive or zero. If they are all zero then empirical data sigmas are estimated by a percentile method in boxes of size given by nsig .
- crval = 1
- Mask value for detected cosmic rays. It is legal for the value to be zero to not mark the cosmic rays in the output mask.
- lcrsig = 10., hcrsig = 5.
- Low and high sigma factors for detecting cosmic rays. These factors multiply the computed or estimated sigma at each pixel and these threshold values are compared to the difference between the candidate pixel and the block average filter value (average of box around the pixel). This only applies to pixels where the block average filter value is within a specified threshold of the background estimate; i.e. the average value is not considered as part of an object.
- crgrow = 0.
- Cosmic ray growing radius. Pixels detected and marked in the output cosmic ray mask by the crval value are increased in size in the mask (but not replaced in the output image) by also flagging all zero valued mask pixels within this specified radius with the cosmic ray mask value. This is done after the detection phase is complete. The separation between pixels is the distance between pixel centers conmputed as a real value. Note a value of at least one is required to affect other mask pixels.
- objval = 0
- Mask value for detected objects. It is legal for the value to be zero to not mark the objects in the output mask.
- lobjsig = 10., hobjsig = 5.
- Low and high sigma factors for detecting objects. These factors multiply the computed or estimated sigma at each pixel and these threshold values are compared to the difference between the block average filter value and the background annulus median. If the values are made very large then object detection can be eliminated and cosmic rays will be detected everywhere.
- objgrow = 0.
- Object detection growing radius. Pixels detected and marked in the output mask by the objval value are increased in size in the mask by also flagging all zero valued mask pixels within this specified radius with the cosmic ray mask value. This is done after the detection phase is complete and so object grown pixels are not used in excluding cosmic ray candidates. The separation between pixels is the distance between pixel centers conmputed as a real value. Note a value of at least one is required to affect other mask pixels.
Craverage detects cosmic rays and objects using a moving block average filter with the central pixel and a specified number of additional highest pixels excluded and a median of an annulus around the block average box. It avoids identification of the cores of objects as cosmic rays by excluding pixels within the detected objects as cosmic ray candidates.
The block average filter computes the average of pixels in a box with the central or target pixel excluded. In addition the nrej parameter can be used to exclude that number of highest remaining pixels as possible contamination from cosmic rays which are larger than one pixel or possibly a very nearby additional cosmic ray. The nrej value should be kept small relative to the total number of pixels in the average so that the average will still be elevated over the median in real underlying objects. The resulting average is used as the prediction for the value of the target pixel. The median of the pixels in a square background annulus around the block average box provides the prediction for the background at the target pixel.
The target pixel is considered part of an object if the difference between the average value and the median background exceeds a specified threshold. If the pixel is NOT considered to be part of an object then if the difference between the pixel value and the average value exceeds a different specified threshold it is identified as a cosmic ray.
The thresholds are defined in terms of sigma factors, which may be different for positive and negative deviations and for object and cosmic ray identification. The sigma factors multiply an estimate for the statistical sigma of the target pixel. The estimate is either based on a noise model or sigma of pixels in a box near the target pixel.
The crmask parameter specifies a pixel mask for the image. If the mask exists then non-zero mask values will be used to exclude pixels from the average, background median, and empirical sigma estimates. Also any pixels with non-zero mask values will not be altered either in the output image or in the final mask. If the mask does not exist then it behaves as if all mask values are zero. If all pixels in the average box or median annulus are previously flagged then the estimates will be undefined and nothing will be done to the output image or mask. Because the task can use an input mask to mark pixels not to be considered it can be used in an iterative fashion.
The noise model is given by the formula
variance = var0 + var1 * data + var2 * data^2
where data is the maximum of zero and the average estimate for the target pixel. The coefficients are all given in terms of the data numbers. This model can be related to common detector parameters. For CCDs var0 is the readout noise expressed as a variance in data numbers and var1 is the inverse gain (DN/electrons). The second order coefficient has the interpretation of flat field introduced variance.
If all the coefficients are zero then an empirical sigma is estimated as follows. The input image is divided into square blocks of size nsig . The (unmasked) pixel values in a block are sorted and the pixel values nearest the 15.9 and 84.1 percentiles are selected. These are the one sigma points in a Gaussian distribution. The sigma estimate is the difference of these two values divided by two. This algorithm is used to avoid contamination of the sigma estimate by the bad pixel values. The block size must be at least 10 pixels in each dimension to provide sufficient pixels for a good estimate of the percentile points. The sigma estimate for a pixel is the sigma from the nearest block. A moving box is not used for reasons of efficiency.
If an output image name is specified then the output image is produced as a copy of the input image but with the identified cosmic ray pixels replaced by the average predicted value. Other optional output images are the average filtered values and the sigma values.
If a mask is specified the detected cosmic rays will be identified with values given by the crval parameter and object pixels will be identifed with values given by the objval parameter. Note that one does not need to use an output image and the cosmic rays can be replaced by interpolation in the data using the tasks crfix , fixpix , or ccdproc .
One final step may be applied to the output mask. The mask values identified with the crval and objval values may be grown by identifying pixel values within a specified radius with the same mask value. Note that this step is done at the end and so any pixels in a preexisting input mask with the same values will also be grown. Also the grown pixels will not affect the output cosmic ray replaced image. See crgrow for a further discussion.
This example illustrates using the craverage task to create a mask with cosmic rays and objects identified and displayed. The image is a CCD image with a readout noise of 5 electrons and a gain of 3 electrons per data number. This implies variance model coefficients of
var0 = (5/3)^2 = 2.78 var1 = 1/3 = 0.34
cl> display obj001 1 # Display in first frame cl> craverage obj001 "" crmask=mask001 var0=2.78 var1=0.34\ >>> crval=1 objval=2 cl> display crobj001 2 overlay=mask001 ocol="1=green,2=red"