Development of laser speckle metrology and its identification techniques

The main goal of this paper is to identify last speckle images rapidly. Digital image processing techniques are employed to analyze the characteristics of laser speckle images and match them up to achieve laser speckle image identification. Besides the database is built to accelerate the identification process and further enhance its practicability. In terms of building the database, Gabor filter is utilized to enhance the extracted characteristics as well as to generate the feature vectors. The final step is adopting K-means clustering to build the classification model of feature vectors. The process of identifying laser speckle images is described as follows. Through experiments we observed that scale invariant feature transform (SIFT) can extract features of laser speckle images very well. However the drawback is that it took too much time to compute and match up those features, which is not suitable for fast laser speckle identification. Therefore the proposed method took enhance SIFT as backbone. Experimental results demonstrate that the retrieval performance of the proposed method is accurate when the database size contains 516 images.