Sonar Image MRF Segmentation Algorithm Based on Texture Feature Vector

• Published in: Global Oceans 2020: Singapore – U.S. Gulf Coast pp. 1 - 6
• Main Authors: Luyuan, Li, Huigang, Wang
• Format: Conference Proceeding
• Language: English
• Published: IEEE 05.10.2020
• Subjects: Feature extraction; fractional differential operation; gray level co-occurrence matrix; Image edge detection; Image segmentation; Image texture; Interference; Markov random fields; side scan sonar image; Sonar; texture image segmentation; Zipf's law
• DOI: 10.1109/IEEECONF38699.2020.9389155

In side scan sonar image processing, the target area and the shadow area are important features of the underwater target. Segmenting it from a background image is an important step in subsequent image recognition. However, the common segmentation algorithms have serious background interference, serious loss of edge contour information, and inconsistent target area and shadow area. Aiming at these shortcomings, this paper proposes an improved Markov random field (MRF) based texture image segmentation algorithm. The feature vector field is used instead of the original MRF observation field, where the feature vectors are obtained by the gray level co-occurrence matrix, the fractional differential operation and the Zipf' s law. In this paper, we compare the results of using three algorithms together and three algorithms respectively on the real side scan sonar image. It turns out that using the three methods together, that is, using the integrated feature vector, a more ideal segmentation effect can be obtained.