can i create feature vector by this way?????

Hi All

i need to ask if it is possible to create a feature vector from the image below by obtaining only the coordinates of white pixels using find(image) and put them in a vector or not ? and then i will create the training set by taking the coodinates of white pixels for 4 similar images ? after that i will perform dimensionality reduction algorithm such as LLE ,laplacian eigenmap ,LDA and matching algorithm ,is this possible or i must use a feature extraction algorithm such as surf or sift?

this is an image:

http://www.2shared.com/photo/zum3JXKy/0015hv1.html

thanks to all