Research Of Face Alignment Based On Simultaneous Inverse Compositional Algorithm

Face recognition,as one of the significant biometric feature technology,has recently been widely studied in recent years,due to its quick and convenient process,visual effects,and simple collection method.Face alignment is the basis of face recognition method,which directly affects the results of face recognition.With further research,face alignment methods are also applied to the face 3D reconstruction,expression recognition,intelligent beauty makeup and therefore have an important significance.In the narrow sense,face alignment refers to eliminating the spatial misalignment,such as transformation,scale changes and rotation.The face alignment in the broad sense refers to automatically locating the key facial feature landmark,such as the points of the eyes,nose,mouth,and contour.Aiming to solve the problem of face alignment,a transform invariant symmetrical principal components analysis(TI-SPCA)framework which aims to automatically align and recognize facial images,and a robust face alignment algorithm based on pose prior are proposed.The main contents are following:1.A transformation invariant symmetrical principal components analysis(TI-SPCA)framework is proposed to automatically align and recognize facial images.Different from the traditional eye-aligned methods,TI-SPCA alignment needs no human intervention.Furthermore,TI-SPCA obtains a transform invariant feature space by minimizing the error between reconstructed images and distorted images.To compare its performance against the eye-aligned method and show its outperformance,this work intuitively demonstrate the visual performance of alignment through the output images of two different alignment methods on ORL database and FERET database.Finally,in order to verify the validity of the aligned image in the recognition system,the proposed method was evaluated by combining three distance functions and four local operators.The experimental results showed the effectiveness of automatic alignment method based on TI-SPCA in face recognition.2.Active Appearance Models are statistical models of shape and appearance,which have shown the capability to achieve exact face alignment.However,it heavily depends on the initialized model and is easily influenced by pose variations,illuminations and occlusions.When the initial model is far from the ground-truth,the performance significantly deteriorates.To avoid such problems,a simple but effective face alignment framework based on pose prior is proposed in this paper.Firstly,we train AAM in-the-wild and use Characteristic Triangle to select a preferable initial model for achieving a robust initialization.Subsequently,we employ a robust and accurate simultaneous Active-Appearance-Model fitting algorithm.Finally,we compare our approach against previous methods and show that it yields the significantly outperformance on challenging LFPW database with large pose variations,occlusions and illuminations.