How do I decide which interpolation method in MOPEX to use to create my science mosaic?
Posted by IRSA HelpDesk on 11 March 2013 01:46 PM
There are 4 interpolation methods available in MOPEX: DEFAULT (1), DRIZZLE (2), GRID (3), and CUBIC (4).

GRID should be used when a quick-look mosaic is needed. It can also be used for background matching ( The gain in speed is up to 10 times as fast as DEFAULT, without much loss in the quality of the interpolated image. The price you pay is the fidelity of the interpolated images. It should not be used for photometry purposes. It should also not be used while doing outlier rejection. If you choose to use GRID, don't forget to set GRID_RATIO in the namelist.

The main differences between the remaining choices are in how the noise properties will be correlated in the final mosaic and in the image quality. Most users will choose between DEFAULT and DRIZZLE. The main advantages of DRIZZLE are reduced noise correlation and marginally improved image quality. The main disadvantages are non-uniform coverage and noise properties which can be difficult to model. CUBIC interpolation could theoretically improve image quality in cases of high S/N because it enforces smoothness in the interpolation function. However, CUBIC interpolation will also produce significantly more noise correlation than a linear interpolation (used in the DEFAULT mode).

A rule of thumb is that if you have a coverage of 10 or less, you will probably want to use the DEFAULT interpolation. The DEFAULT interpolation method has one optional input parameter: Fine Res (FINERES in namelists). This should be set to zero, if present at all, as this feature is not fully implemented. If you have a coverage of 10 or more, you may want to consider the use of DRIZZLE. The only input parameter for this is the Drizzle Factor (DRIZ_FAC in namelists). This is the factor by which each input pixel is shrunk before it is projected onto the final image. The smaller you make the Drizzle Factor, the more coverage you need. If you use DRIZZLE with a small Drizzle Factor and/or low coverage, you could produce holes in your mosaic. Even if you don't end up with holes, the final pixels are more likely to be affected by outliers because fewer data pixels go into each mosaic pixel. The coverage (and hence effective gain) of the mosaic could show significant variations from pixel to pixel.

More information on image interpolation can be found in the MOPEX User's Guide.
(0 vote(s))
This article was helpful
This article was not helpful

Comments (0)