| 摘要 |
The present invention relates to a medical image search method based on image clustering. The method includes: (a) a step of dividing each of a plurality of inputted medical images into sub areas and extracting an oriented center symmetric local binary patterns (OCS-LBP) feature value by patch unit; (b) a step of clustering the feature value into a code book and converting each image divided by patch unit into a bag-of-feature (BoF) through the code book; (c) a step of generating the BoF feature as a feature vector and learning and classifying the feature vector through a random forest learning classifier; (d) a step of repeating (a) through (c) when a query image is inputted and measuring the similarities with the medical image through a small number of specific classes among the classified classes; and (e) a step of outputting a search output image in accordance to the measured similarity. The present invention provides the medical image search method suggesting a new OCS-LBP feature suitable for a medical image to efficiently search for the medical image, applying BoF to efficiently reduce a feature dimension, and show a high performance search result by measuring a similarity distance through the random forest classifier to improve the search speed and performance instead of matching the query image with the entire image. |
