Band-limited Imaging with Undersampled Detectors

Over the past decade “Drizzle” has become a de facto standard for the combination of HST images. However, the drizzle algorithm was developed with small, faint, partially-resolved sources in mind, and is not the best possible algorithm for high signal-to-noise unresolved objects. Here, a new method for creating bandlimited images from undersampled data is presented. The method uses a drizzled image as a first order approximation and then rapidly converges toward a bandlimited image which fits the data given the statistical weighting provided by the drizzled image. The method, named iDrizzle, for iterative Drizzle, eliminates the small high-frequency artifacts that can be introduced by drizzling. The method works well in the presence of geometric distortion, and can easily handle cosmic rays, bad pixels, or other missing data. It can combine images taken with random dithers, though the number of dithers required to obtain a good final image depends in part on the quality of the dither placements.