Exoplanet direct imaging involves very low signal-to-noise ratio data that need to be carefully acquired and
processed. This paper deals with data processing for the VLT planet finder SPHERE, that will include extreme
adaptive optics and high-contrast coronagraphy, and where field rotation will occur. First, we propose estimators
of the intensity, the intensity estimate uncertainty, and the initial position of a potential exoplanet. Because of
the very large amount of data to process, they are derived from a simple gaussian data model relying on the
time-stationarity of the background, where the so-called background is everything but the exoplanet. Analytical
properties of the estimators are given, under the gaussian data model and under a more sophisticated data model.
Then, in order to relate the detection procedure to a probabillity of false alarm, the detection consists simply
in thresholding the intensity estimate at a given initial position. Finally, this detection-estimation algorithm is
applied on a dataset simulated using the CAOS-based Software Package SPHERE, including time evolution of
the atmospheric, pre-, and post-coronagraphic quasi-static aberrations. As a preliminary result, the detectionestimation
algorithm proves to be totally satisfactory for a 8 × 10-5 intensity ratio for exoplanets located from
0".2 to 2". The stationarity assumption is discussed.