Recognition of human kidney ultrasonic organ images by two‐dimensional dp matching method

This paper discusses the application of two-dimensional dynamic programming (DP) matching to ultrasonic kidney images. By introducing the model-matching technique, feature extraction of the object organ is achieved. It is also possible to reconstruct the three-dimensional structure of the organ. The proposed technique has the following features. (1) Unlike the monotone continuous condition employed in speech recognition, the two-dimensional DP method has no time-sequence constraints. (2) The recognition performance is improved by utilizing the edge information. (3) Hierarchical construction of DP from coarse to fine grids improves the processing speed. (4) The method allows man-machine interaction by fixing a point on the contour of the object. (5) The method can be applied to time-series and/or three-dimensional consecutive images and the speed is improved by using the object from the previous image.