Indian Journal of Science and Technology
DOI: 10.17485/ijst/2015/v8i31/78429
Year: 2015, Volume: 8, Issue: 31, Pages: 1-3
Original Article
Asma Lak1*, Naser Parhizgar2 and Mahbobeh Lak1
1 Department of Electrical Engineering, College of Engineering, Geneva branch, Islamic Azad University, Geneva, Iran
2 Department of Electrical Engineering, College of Engineering, Shiraz branch, Islamic Azad University, Shiraz, Iran
In this paper a new method has been proposed to less time (by reducing the repetition rate) and more accuracy in EM algorithm running. In this paper MATLAB software has been used to simulation. The image which has been used to simulation taken by AVIRIS in June 1992 in Indiana with 145*145 pixels. EM algorithm is an iterative algorithm and uses from the results of previous step to re-estimate. This repetition is continued till the stop condition is satisfied. The proposed method uses from the probability of belonging of each pixel to a class and compares it with estimated probability of previous stage. After running proposed algorithm the time of running and the accuracy have been improved. The time has been reduced from 1370 to 1581 seconds because the number of iterations has been reduced from 14 to 12. The accuracy increased from 66.757 to 69.979. This algorithm is useable to classify the images which need to more accuracy.
Keywords: Accuracy, EM Algorithm Image Classification, Improved Algorithm, Running Time
Subscribe now for latest articles and news.
