Abstract: This paper proposes an iris feature extraction algorithm based on energy compensation and feature weighting to solve the problems that the capacities of the feature extraction with different sub-band of the Contourlet transformation are different. It uses orthogonal images to achieve energy compensation for the original images, estimates the weights of sub-bands by using General Gaussian Distribution(GGD), and gives larger weight for the feature with better classification capacity, so that the statistical information of the samples is made full use of and the features are extracted efficiently. Experimental results show that the algorithm has good robustness and improves iris recognition rate.

Key words: Contourlet transformation, iris feature extraction, energy compensation, feature weighting, General Gaussian Distribution(GGD)

摘要： Contourlet变换不同子带的特征提取能力存在差异。针对该问题，提出一种基于能量补偿和特征加权的虹膜特征提取算法。采用正交图像对原图像进行能量补偿，利用广义高斯分布估计各子带数据的权值，为分类能力强的特征量赋予较大权值，以充分使用样本的统计信息高效地提取特征。实验结果表明，该算法的虹膜识别率较高，鲁棒性较强。

关键词: Contourlet变换, 虹膜特征提取, 能量补偿, 特征加权, 广义高斯分布

CLC Number: 

• TP301.6