What does SIFT do? does it establish the feature correspondences between views? Does it receive the corners of the image sand remove the ones which are not descriptive? Does it do the matching?
SIFT is a short cut for "Scale Invariant Feature Transform". SIFT features are often used to locate descriptive points in images. For image stitching SIFT features are located in two images. Then you look for similar points / features in both images. If enough correspondences are found and the images contain some common part.
SIFT features have a location in an image, in addition the feature contains a rough description of nearby area of the image (the size of this area depends on the scale). This description is used test if two SIFT features are describing the same location.