To improve the measurement angular resolution of star sensors while maintaining high update rates, the detector of four million pixels, parallel output data through multi-channels is chosen. The problem of star sensor’s image processing based on this kind of detector is solved by designing a method of using marker to achieve extraction of star centroid fast and efficiently. Algorithm is divided into markers calculation of multi-channel and target processing of channel boundary：markers calculation of multi-channel is mainly to solve the problem of image processing within the channel, while among channels using parallel processing method; target processing of channel boundary is mainly to solve the target cross border splicing problem. In the high frequency clock is 100 MHz, under the condition of using the FPGA implementation algorithm based on level 4 lines parallel processing architecture, and the actual test validation. The validation results show that the algorithm can quickly and correctly extracts as target point on the surface of the detector. Accuracy of positioning extraction of single star is better than 1/172 pixel, update rates is better than 9.5 Hz, which meets the requirement of the star sensor accuracy and high update rates.