Detection of microaneurysms in bifrequency space based on SVM

The early detection of microaneurysms is important for tumor diagnosis and treatment, while it is difficult to find via naked eyes. This paper proposed a method to obtain microaneurysms in bifrequency space based on SVM. Firstly, a generalize histogram algorithms was employed to enhance the images, which achieved a well SNR. Secondly, the grayscale image is subjected to the radon transform and then it is subjected to HOS to extract the bispectral invariant features, which construct the bifrequency space Finally, the features were extracted and inputted to the SVM to classify the microaneurysms. The experimental results showed that proposed method achieved an accuracy of 94%.