Abstract. In this study, we analyze a series of images of Io obtained with theEuropean Southern Observatory adaptive optics system (Adaptive Optics Near Infrared System, ADONIS) at 3.8 tzrn from 1996 to 1999, with particular emphasis on the observations carried out in late 1999 in support of the Galileo fiybys of Io. Use of a new myopic deconvolution method, Myopic Iterative Step Preserving Algorithm (MISTRAL), especially designed for planetary objects, significantly improves the quality and the reliability of the reconstructed images. Using a simulation of artificial images of Io, we estimate the extent to which this algorithm is able to prevent noise amplification, to better restore sharp edges, and to preserve the initial photometry. Once this a•lgorithm has been applied to our data and the solar-reflected background has been subtracted, we have 89 images of Io available for search and temporal survey of the hot spots over 4 years. In most cases, the data, acquired during two consecutive nights, provide an almost global view of the surface of Io. We identify 20 hot spots for which we determine the coordinates and the brightness at 3.8 tzrn (as a function of time). More than half of the hot spots were detected on all the images and were considered as persistent. Particular emphasis has been put on the brightest two hot spots, Loki and Pele, for which we obtain a center-to-limb variation curve (a cosine curve for Loki and a curve decreasing slightly faster than a cosine law for Pele with possible physical interpretations).