On the basis of aerial photos of sea ice in the Second Chinese National Arctic Research Expdition during July and September 2003 in the area of 74.11°~79.56°N and 144.17°~169.95°W, image processing techniques are used to acquire some geometric parameters of floes, such as fractal dimension, roundness and mean caliper diameter. Latitudinal variation of morphology of arctic ice floes is then obtained, after comparing fractal dimension and roundness which represent geometry of floe edge, the latter is found to show a more obvious tendency with latitude but still not enough to describe floe abrasion. Then mean caliper diameter of ice floe is used as characteristic size to analyze floe size distribution, the result reveals that cumulative probabilities of floe sizes agree well with a power-law function, and distribution dimension is generally in the range of 1.05~1.25 and slightly increases as the latitude increases. However slight curvatures are still observed in the plots of cumulative probabilities as in former researches, which can be attributed to both the limitation of sampling area and effect of thermodynamic process. On the basis of aerial photos of sea ice in the Second Chinese National Arctic Research Expdition during July and September 2003 in the area of ​​74.11 ° -79.56 ° N and 144.17 ° -169.95 ° W, image processing techniques are used to acquire some geometric parameters of floes, such as fractal dimension, roundness and mean caliper diameter. Latitudinal variation of morphology of arctic ice floes is then obtained, after comparing fractal dimension and roundness which represent geometry of floe edge, the latter is found to show a more obvious tendency with then mean caliper diameter of ice floe is used as characteristic size to analyze floe size distribution, the result reveals that cumulative probabilities of floe sizes agree well with a power-law function, and distribution dimension is generally in the range of 1.05 ~ 1.25 and slightly increases as the latitude increases. However slight curvatures are still observed in the plots of cumulativ e probabilities as in former researches, which can be attributed to both the limitation of sampling area and effect of thermodynamic process.