Under the influence of reconfigurable embedded computing techniques, advanced reliable space computer has been a vital research area for several years. This paper presents a FPGA-based architecture for run-time parallel processing on space solar telescope (SST), a scientific solar-observation satellite. SST is required to process onboard a huge amount of image data observed through multi-channel CCD cameras- around 1728 GB per day, which requires processing speed more than 10,000 MIPS, if an instruction-set-based processor is adopted. Thus, a FPGA-based reconfigurable architecture is proposed to construct SST's computing core for the purpose of multi-channel parallelization and self-healing capability, when running in severely-radiate solar obit. That is, partial reconfiguration can help "heal" single-particle upset errors imposed by space radiation. Our space reconfigurable specimen machine is composed of commercial (off-the-shelf) Xilinx FPGAs (XC2V1000s and XC2V 3000) and 2GB external Flash-RAMs. In general, the whole processing system is a combination of partial reconfigurable DSP clusters and an embedded LEON2 processor, targeting high-performance payload computing and data transmission in outer space; three reconfiguration strategies are utilized to guarantee system reliability and flexibility.