Information of cellular organelles location and morphology is essential for cancer simulation. In order to obtain such information, the segmentation of the organelles from electronic microscopy intracellular image is crucial. In this paper, we focus on the automatic segmentation of mitochondria organelle which is one of the most important organelles tightly related to the form of cancer. A simple three-stage strategy which includes coarse segmentation, detection and fine segmentation is proposed for fully automatic mitochondria segmentation. The local gradient calculation provides a weight factor matrix and a orientation matrix. The weight factor matrix will improve the contrast of organelle boundary of intracellular images and hence facilitate both coarse and fine mitochondria segmentation. The orientation matrix will be used for enhancing the Gabor feature extraction which make the mitochodrial detection process more accurate. Machine learning-based classifiers including k-nearest neighbor (k-NN), support vector machine (SVM)-based and neural network (NN)-based classifiers, are considered to learn with eight extracted features for mitochondrial detection. Experimental results on focused ion beam (FIB) and scanning electron microscope (SEM) images of cancer cellular of human head and neck squamous cell carcinoma (SCC-61) have shown the effectiveness of proposed method.