Plant identification is a vital task in areas such as medicine, botany and food sector. Leaves also play an essential role in plant species recognition. However, plant identification becomes more challenging in case of leaves with the complicated background having interferences and overlapping. In this paper, an efficient approach is proposed for classification of leaf images with a complicated background. In the proposed approach, firstly, a saliency detection method is used to segment leaf images from a complicated background. Then, morphological and texture features are extracted from the region of interest. Finally, a multilevel classification is used for Leaf recognition. The experimental results are validated with an accuracy of 94%.