With the increase of ball firing times, multiple flaws will appear in the gun bore, due to the impact of high temperature-high pressure powder gas and the attrition of pills, which influences the fire accuracy and the fire safety. Through analyzing the gradient and direction characteristics of indentation, rust corrosion and burning corrosion, the gun bore flaw images are separated into spread flaw images and regional flaw images. Then the classification factor of the spread flaw image and the regional flaw image is defined according to the sensitivity to direction of Radon image transformation and the edge gradient of Susan operation. This classification offers basis for the following analysis. Experimental results show that the algorithm could classify the gun bore flaw image accurately through computing the flaw characteristics and thus realize the quantitative classification.