Multi-class Support Vector Machine classifier (M-SVM) method is proposed in this paper to achieve high accuracy, MCC, Sensitivity, specificity, and F-score. The Fuzzy C Mean clustering and OTSU threshold value are used for segmentation, Gray Level Co-occurrence Matrix is used to extract better features after elimination of noise from satellite images by using Dual tree complex discrete wavelet transform (DTCWT). The extracted features are exposed to the infinite feature selection algorithm in order to recognize the required arrangement of elements to the proposed multiclass SVM algorithm for ice type classification. Advanced Microwave Scanning Radiometer - Earth Observing System (AMSR-E) images are considered as test images for the proposed method. The effectiveness of M-SVM is compared with Randem Forest (RF), Neural Network (NN) and other methods. From the findings, the proposed method increases accuracy, MCC, Sensitivity, specificity, and F score by 1.90%, 3.06%, 0.81%, 2.40%, and 2.53% respectively as compared with RF method; 9.0%, 23.62%, 2.54%, 11.37%, and 11.75% as compared with NN method respectively.