Abstract: It is well-known that the low-level features can hardly express high-level semantics, which is also known as Semantic Gap.In order to resolve the barrier of semantic gap, this paper proposes an approach to Chinese paper-cut recognition by the feature combination and selection with spatial-constraints.This approach first extracts shape features of the paper-cut by the integration of spatial pyramid matching model and contextualizing histogram model to obtain spatial features of the paper-cut;then the feature combination and selection are conducted to get discriminative features of the paper-cut by AdaBoost;at last, the recognition of the paper-cut can be achieved.