The　traditional　eigentransformation　method　for　face　hallucination　is　a　linear　subspace　approach,　which　represents　an　image　as　a　linear　combination　of　training　samples.　Consequently,　those　novel　facial　appearances　not　included　in　the　training　samples　cannot　be　super-resolved　properly.　In　this　paper,　a　KPLS　(Kernel　partial　least　squares)　regression　is　introduced　into　the　eigentransformation　method　to　reconstruct　the　High　resolution　(HR)　image　from　a　Low　resolution　(LR)　facial　image.　We　have　evaluated　our　proposed　method　using　different　zooming　factors　and　compared　these　performances　with　the　current　Super　resolution　(SR)　algorithms.　Experimental　results　show　that　our　algorithm　can　produce　better　HR　face　images　than　the　compared　eigentransformation　based　method　and　the　KPLS　method　in　terms　of　both　visual　quality　and　numerical　errors.