A new active contour model based on the Conscience, Archiving and Mean-Movement mechanisms and the SOM

Active contour models are widely used in extracting object boundaries. However, most of these models usually fail to capture concave boundaries properly and impose high computational cost. In this paper, a new active contour model based on the Conscience, Archiving and Mean-Movement mechanisms and the SOM (CAMSOM) is proposed to eliminate these deficiencies. The proposed method extends the Batch SOM method (BSOM) by introducing three mechanisms of Conscience, Archiving and Mean-Movement mechanisms. To evaluate the performance of the proposed method compared with both energy minimization and SOM-based methods, some experiments are carried out on a set of grayscale images including synthetic and real ones. The experimental results are compared with those of the BSOM in terms of accuracy and convergence speed. The results reveal that, compared to BSOM, the proposed method requires less computations to converge to the object boundaries and extracts the boundaries of complex objects more accurately, even in the presence of weak or broken edges. Moreover, CAMSOM has higher performance and accuracy in capturing the boundaries of the objects placed arbitrarily in a multi-object scene, whereas the performance of BSOM in multi-object scenes highly depends on the arrangement of the objects. Compared to the energy minimization methods, the proposed method can accurately and quickly converges to the concave boundaries, whereas gradient vector flow (GVF) and vector field convolution (VFC) which are two well-known energy minimization methods get stuck in local minima and cannot proceed to the end of the concavity.

► We propose an improved SOM-based active contour model named CAMSOM.
► CAMSOM incorporates Conscience, Archiving and Mean-Movement mechanisms.
► CAMSOM performed better compared with SOM-based and energy minimization methods.
► CAMSOM can capture long and thin concave boundaries accurately and quickly.
► CAMSOM can find boundaries of the objects placed arbitrarily in a multi-object scene.