On a grism image where many objects are detected, it is likely that the spectrum of one object may overlap with the spectra of others. In these cases, attempts can be made to "deblend" the spectra; i.e. remove or minimize contamination of one spectrum from another.
First, all spectra that overlap are identified and categorized based on the closeness and size of the objects according to three different (user definable) thresholds. The results of this categorization determines whether it is possible (or worth while) to attempt to deblend the spectra.
When it is impossible to remove spectral contamination, the wavelengths affected are noted in the catalogue entry and on the hardcopy output. If there is only a slight contamination, a warning is issued. These two features will be implemented in future development.
In cases where it is possible to deblend a particular spectrum it is necessary to determine which other spectra are affecting it and to calculate the contamination factors at each wavelength. The contamination factors are computed by summing the weights corresponding to the portion of overlap. For example, if object J is contaminating object I's spectrum, the contamination of I is merely the sum of object J's weight in the zone of overlap. This sum is in the y-direction at each wavelength affected. This process of summing the weights is performed for each and every spectrum affecting another object's spectrum.
Once all contamination factors are calculated they are assembled into a square matrix where the i-th column in row j contains the contamination of spectrum i from spectrum j. One matrix is created for each wavelength in the spectrum. In order to correct the spectra, we simply perform a matrix inversion on each matrix. The elements of these inverted matrices are used to calculate a correction coefficient for each flux value of the contaminated spectra.
A possible scenario is that two or more spectra are almost aligned and separating them is quite impossible. This situation shows itself in the form of a singular matrix. In future releases of Calnic C, such cases will be noted in the catalogue and log file and attempts will be made to find an invertible submatrix in order to deblend other objects' spectra.