It takes a long time for high speed frame cameras to record image information of target states. The images need to be compressed, recorded and transmitted in real time for limited storage capacity of record medium. The speed of EZW and SPIHT coding methods are both too slow to work out on field-programmable gate array (FPGA) and they need high quality buffer storage. A new coding algorithm has been proposed based on redundancy and distribution characteristics of wavelet coefficients after wavelet transformation. Wavelet systems are grouped dynamically by directly comparing with magnitude of coefficients and incorporating characteristics of numbers and distributions. Less memory space is used to record coefficient information of images. The algorithm is suitable for FPGA. The feasibility of the method is verified by experiments.