A new set of recently developed LAI algorithms is employed for a production of a global LAI dataset covering the whole year 2003 in 1 km resolution and with a time-step of 10 days, using SPOT-4 VEGETATION data. An automated erratic cubic-spline capping method is used for reconstructing the seasonal trajectories of LAI. The method is proved to be effective in substantially improving the quality of the dataset by limiting the effect of erratic values, present in the original dataset. The validation of the final LAI product is made by comparing the result with a set of reference sites from Canada and China, covering a wide spectrum of land cover conditions. A procedure for correct geo-referencing and positioning of corresponding scenes that employs point spread functions and significantly improves the quality of the match of the corresponding scenes is developed. The global dataset is found to be reasonably accurate in retrieving the real LAI values and superior to global MODIS LAI product; sources of potential errors are subsequently identified and explained.