The objective of this paper is to propose a novel approach to recognize Characters (Digits, Upper case letters and Lower case letters) of text spotted in natural scene image with cluttered backgrounds. The projection profile method is used to segment localized text component into individual characters, which is then submitted to character recognition process. In the proposed methodology, the Moment Invariants (MI) features are extracted from the Contourlet transformed image of segmented input characters. The PNN classifier is used for character recognition and its performance is compared with KNN classifier. The testing has been done using images taken from ICDAR dataset and personal dataset. The performance is measured in terms of recognition accuracy. The experimental results demonstrated the efficiency of the proposed method