| Face recognition technology is widely used in security, offender tracking and authentication. Especially in the last twenty five years, both of face detection technology and face recognition technology are gained a huge of attention and development. Now face recognition not only applied for the simple background, expressionless image and nicely input,but used effectively in complex background, expression and posture change and non-frontal face images exist. With the rapidly development of computer hardware and computing power,people are not satisfied with study the face recognition in two-dimension, but in the three-dimension.SPM+ which is based on 2DPCA algorithm. The algorithm obtains the distribution of the size of the face image information by calculating the entropy of the image. According to the effects that we gain, we give different weights to calculate the face recognition. The difference between the original module-based 2DPCA algorithm and the new one is that the original one divides the image by the size of the image, but the new one divides the image by the weight of the information. Then we use MSD algorithm which is used to study the difference between people-face image and background image. At last we combine SVM algorithm in the methods which has excellent performance of research in binary classification problem. From the experimental, we found that the SPM+ algorithm, whether in perfect situations or in a complex situation shows good performance than other traditional algorithms.This article also built low-resolution, high noise face database which is mostly like the normal situation. The SPM+ is still better than others under the above conditions.In the pre-processing stage, we use ? method to reduce noise. We can compare median filter and mean filter under Gaussian noise and salt and pepper noise. In order to modify the original image in a geometric level, we detect human eyes with AdaBoost algorithm after detect human face based on AdaBoost algorithms. |
PDF Full Text Request