The quality of crop yield is reduced due to leaf diseases in agriculture. Therefore, it is possible to automate the recognition of leaf diseases to improve yield in the farming sector. However, most systems lack in performance due to different patterns of leaf disease that influence the precision of detection. In this paper, a computer vision framework is developed by framing a model that consists of image acquisition, feature extraction and image classification. A deep learning classifier namely Deep Belief Network (DBN) is used for classification of realtime images. The experimental results on pepper plant leaf disease detection show that the proposed method has improved rate of classification than other existing methods. The classification result shows that whether the leaf is diseased or not.