Non-stationary nature of SAR images is a major obstacle to the automatic interpretation of SAR sea ice image. Incidence angle effect is one of the main factors leading to instability in the sea ice image features. Based on Radarsat-1 ScanSAR mode data

this paper proposes a new segmentation algorithm which integrates incidence angle effect correction step. Considering both the speckle noise and incidence angle effect

the regional clustering

incidence angle effect correction and region merging will be combined through the pathway starting from the pixel

region and to the large-scale area. The efficiency of the proposed method has been demonstrated on the segmentation of Baffin Bay SAR sea ice image and Gulf of Bothnia SAR sea ice image

suggesting that the segmentation accuracy has been substantially improved in contrast to area-based MRF algorithm.