Least squares image estimation in the presence of drift and pixel noise

We discuss Least Squares (LS) image estimation for large data in the presence of electronic noise and drift. We introduce a data model where, in addition to the electronic noise and drift, also an additional type of noise, termed pixel noise, is considered. This noise arises when the sampling does not take place on a regular grid and may bias the estimate if not accounted for. Based on the model, we present an efficient Alternating Least Squares (ALS) algorithm, producing the LS image estimate. Finally, we apply the ALS to the data of the Photodetector Array Camera and Spectrometer (PACS), which is an infrared photometer onboard the European Space Agency (ESA) Herschel space telescope. In this context, we discuss the ALS implementation and complexity and present an example of the results.