Support vector machine (SVM) hyper plane formula

Question

I'm working on image classification system, i used sift and support vector machine. I want to make it a web application using java applet, How can i use SVM results to classify new images? Is there a formula for svm hyper plane that i can use? or anything else can be used to classify. actually i calculated the sift for all images and i want to have a web application that a user can browse an image from predefined images (which it's sift is calculated) and classify to see the results. PS: i did the training phase using MATLAB i just want the final decision for classification of new images.

Answer 1

Just a basic thought - if you already have predefined images, with precalculated sifts, why not preclassify them too?

The classification "formula" for SVM can be of two types:

a) for linear case, assuming that you have w (features weight vector) and b (threshold)

f(x)=sgn( w'x + b )

b) for kernelized case, assuming that you have alpha (dual coeffficients), sv (support vectors), K (used kernel function) and b (threshold)

f(x)=sgn( sum_i alpha_i K(sv_i,x) + b )