I'm trying to classify a set of binary images in matlab. I have created a model using a lot of different images and have classified each one accordingly so that the model 'knows' which images belong to which class. Each image is represent as a feature vector, which is 1x59. I am successfully able to calculate the mean and covariance of class of images. The classifier uses maximum a posterior classification (MAP), but for some reason all the images in the test aren't being classified properly. The long equation I am using in the variable 'score' is the MAP calculation . Does anyone have any suggestions as to how to best classify an image?
function [classname] = classify(imagepath)
load('models');
N = length(models(1).mean);
features = getFeatures(imagepath, ((N+1)/2));
maxscore = -inf;
for index = 1:length(models)
model = models(index);
score = max(log(model.prior) - 0.5*log(abs(model.cov)) - 0.5*(features - model.mean)'*inv(model.cov)*(features - model.mean));
if score > maxscore
maxscore = score;
bestindex = index;
end
end
classname = classes(bestindex);
end
