For incorrect segmentation caused by the lack of boundary information, a region-based multiscale segmentation method of panchromatic remote sensing image is proposed. Firstly, curvelet transform based multiscale analysis is used to obtain the multiscale decomposed image. Then its image domain is partitioned into a set of blocks by regular tessellation, and a multiscale statistics based segmentation model united with region is built by using the statistic method. Furthermove, a reversible jump Markov chain Monte Carle(RJMCMC) algorithm is designed to simulate the segmentation model. The proposed approach is used to segment panchromatic remote sensing images, and the results show that the proposed approach can effectively solve incorrect segmentation and segment the image well.