| Abstract
|
:
|
w etheGood face detection, is an importmt part of a reliahle face recogn,ition, system,.propose methods of face detection, and n,orm,alization,. Normalization, req~l,iresexact location, of eyes t o stmighten, any rotated face image. Eyes detection, m,eth-ods are also developed for this purpose. To im,prove the accuracy of the classicalPCA techn,iqu,e for face recogn,ition,, we delieloped a system hased on, the method ofleast-squares which takes in,to accomt the class in,formation,. W e further improveaccuracy hy vsin,g m~~,ltiple scales of the sam,e face image. Such scales are ohtain,edusing the discrete wavelet tran,sform. W e shouc that there exists a particdar set ofsuhhand images that are essen,tial for roh& face recogn,ition,. W e propose here am,odel t o find these svhhan,ds for in,put im,ages. Based on, this, we develop a mul-tiscale face recogn,ition system which ~l,tilizes the m,odel t o find importan,t subhan,ds.These svhhan,ds are then, used for im,provin,g recogn,ition accwacy..
|