7.2 Combining more than two images (SUMIM, SPIXR, STACK)

The task SUMIM is used to sum or average any number of images. Since AIPS has only a limited number of adverbs of the kind INNAME, IN2NAME, etc., SUMIM requires that all input images have identical INNAME and INCLASS. The input images are then specified by INSEQ (the sequence number of the first input image), IN2SEQ (the sequence number of the last input image), and IN3SEQ (the increment in sequence number). All input images have to reside on the same disk.


to review the required inputs.

> INDI 0 ; MCAT  C R

to help you find the catalog number of the first input image.

> INDI n ; GETN ctn  C R

to select the first input image from disk n catalog slot ctn.

> IN2SEQ s  C R

to specify the sequence number of the last image to be included.

> IN3SEQ 0

to specify the increment in sequence number (= 1).

> OUTN xxxxx  C R

to specify xxxxx for the name of the output image.

> OUTC ccc  C R

to specify ccc for the class of the output image.


to specify the factor with which to multiply each image before adding. f = 1 leads to summation, f = 0 defaults to the inverse of the number of input images (average)

> GO  C R

to start SUMIM.

This is a very noisy way to make a line-sum image. For more serious work, use BLANK (7.4) and XMOM (8.6) instead.

A popular analysis technique has been developed in which images centered upon objects of some chosen kind are simply added up even though the objects’ emission is less than the noise. The belief is that, if there is real emission just below detectability in individual images, the summed image will reduce the noise and reveal the weak emission. In 31DEC15, the task STACK attempts such an analysis doing a weighted mean or median of a matched set of images ignoring the coordinates.

The task SPIXR is intended to fit spectral indexes optionally including curvature to a cube of image planes. This cube is build with FQUBE to make an FQID axis out of irregularly spaced frequencies. The cube is transposed by TRANS to put the FQID or, if build with MCUBE, frequency axis first. Then SPIXR will do a least squares fit for spectral index. Again, the results will be noisy unless the initial images have been blanked and converted to similar spatial resolution.