how I can extract quadratic features from surf features in Matlab for face recognition project.

I am doing a face recognition project based on matching surf features.I tried to apply gabor filters to improve results but they did not work now i want to extract quadratic features from surf features so system can match features based on euclidean distance. Can anybody help me with that or suggest something to improve results.

query.wholeImage = imread('F:\images\test.jpg')
  faceDetector = vision.CascadeObjectDetector();
   bbox  = step(faceDetector, query.wholeImage)
query.image=imcrop(query.wholeImage,bbox(1,:));
query.image=rgb2gray(query.image);
% Detect SURF features
query.points = detectSURFFeatures(query.image);
% Extract SURF descriptors
 [query.featureVectors,query.points]=extractFeatures(query.image,query.points,'SURFSize',64);
%imageCollection has feature surf features data set of train images   
for i=1:numel(imageCollection)
 [matchesind, distanceind] = knnsearch(imageCollection(i).featureVectors,query.featureVectors,'Distance','euclidean','K',1);
 distance(i)=sum(distanceind(:));
end
[m mi]=min(distance);
matched=imshow(srcFiles(mi).name);

[Dima Lisin]

Could you explain what quadratic features are?