Image processing method for positioning eyes

Abstract

The invention relates to an image processing method for positioning eyes. The invention adopts a technical proposal that the method comprises the following steps of: acquiring an image of a face; and detecting the image of the face, establishing probability support vector machine appearance models for three components including eyes and nose respectively, establishing the Gaussian distribution models of the measuring parameters of the spatial structure of the three components respectively, redefining the length of three edges and establishing Gaussian distribution models, establishing an enhanced image model, selecting candidate points of maximum appearance characteristic similarity and maximum structural similarity as the optimal positions of a target, and determining the possibility of human eyes z1 and z2 or nose z3 according to an appearance characteristic ui shown by the probability support vector machine appearance models on the basis of the probability support vector machine appearance models to position the eyes. The method overcomes the drawbacks in the prior art such as high requirements on image resolution ratio, light and definition, and difficulty in the positioning of the human eyes in a complex environment accurately. The method establishes a high robustness space structural model starting from the appearance characteristics of eyes and the characteristics of the spatial distribution structure of the eyes and the nose and improves human eye positioning accuracy.

Description

technical field [0001] The invention belongs to a biological feature recognition method in the fields of pattern recognition and computer vision, in particular to a robust and accurate human eye positioning technology under uncontrolled conditions. Background technique [0002] As a new biometric authentication method, face recognition has attracted much attention due to its potential wide application prospects in the fields of public security, information security, and finance. A complete face image information processing system consists of the following key modules: face image acquisition, face feature point location, face normalization, face feature extraction and feature matching recognition. [0003] Regularized faces are an important condition for the execution of face recognition algorithms, which can greatly improve the reliability of face recognition. However, face detection can only obtain the approximate area and size of the face, so the face detected by the face...

AI Technical Summary

Problems solved by technology

However, in this model, the simple Gaussian distribution appearance model is not enough to describe the complex changing mode of the human eye, and the structural model based on the relative position distribution between features is not robust to rotation, scaling and translation transformation, so it is difficult to realize the complex environment. precise positioning of the human eye

[0009] To sum up, in the real environment, the appearance of the human eye is complex, and it is difficult to achieve accurate human eye positioning only based on the appearance features or combined with a simple spatial structure model

Images