This work proposes a methodology for detecting pomegranate leaf diseases early and accurately using image processing techniques and Support Vector Machine (SVM). Color image segmentation using K-means clustering technique is performed to extract the region of interest from the pomegranate leaf image. Further significant texture and color features are extracted from the region of interest for the purpose of training SVM classifier. Classification is performed by considering two different feature sets viz. i) entropy and saturation ii) hue and energy. Experimental results show that SVM classification is highly accurate with entropy and saturation feature set compared to that of energy and hue set. This automated system assists farmers to detect the healthy & diseased leaves without human intervention.