According to the characteristics of high/low scan frequency and dense/sparse scan tracks in central/fringing area of field of view of rosette scanning, a novel unequally spaced sampling method be used in rosette scanning infrared imaging subsystem is proposed to meet the reguirements of short processing time and good image quality. The experimental results show that the optimum unequally spaced sampling function is gaussian. The relations among image data quantity, resolution and unequally spaced sampling parameters are discussed in detail and the method of parameter determination is consequently given.