How to apply sliding window for car detection in HOG+SVM ??

Dear All, I have trained the SVM classifier with HOG features of positive (car) and negative (non car) samples. Now in the detection part, I have tried to use blockproc for sliding windows. The code is pasted below:

fun = @(block_struct) hog_feature_vector(block_struct.data); out = blockproc(double(Img),[64 64],fun);

Out contains a feature vector of size 4x7056 for the test image of size 256*256. Now how to detect only those regions which contain a car? Please give me some ideas.

Thanks in advance.