Improved Curvature Scale Space Based on Image Retrieval Using Dynamic Time Warping

摘要

In this paper, we suggest improved Improved Curvature Scale Space(ICSS) based on image retrieval using Dynamic Time Warping(DTW). This method consists of three process. At the first stage, we retrieved an image using ICSS method. We applied ICSS space to extract shape information. This method expressed contours of the object which is binarized through pre-process in ICSS space, then extract shape features of the object. At the second stage, we applied DTW for time series to determine similarities. DTW algorithm is good to find the optimal alignment between two time series. finally, merge ICSS shape feature with DTW information. This method is estimated with a database of 1000 images. With estimated result, we show that proposed method is better in accuracy of searching than without DTW.