Image Segmentation Method With Improved GA Optimization of Two-Dimensional Maximum Entropy

Image segmentation is the cornerstone that determines the effectiveness of image processing, but traditional image segmentation methods have issues such as long computation time, low recognition accuracy, and poor anti-interference ability. To address this issue, research improves the genetic algorithm using adaptive hybridization and adaptive mutation probability, and combines it with the Bat algorithm to optimize the local optimization problem of the image. The sparrow algorithm is utilized to optimize the two-dimensional maximum entropy of the image, and the nonlinear inertia weight factor is brought to optimize the local search ability. The Levy flight constant is used to overcome the local optimization problem. The experiment findings indicate that the optimized algorithm improves the similarity of medical image features by an average of 11.2%, reduces segmentation accuracy by 2.6% under noise interference compared to other algorithms, and has an average peak signal-to-noise ratio 0.96 higher than other algorithms. From this, the improved algorithm greatly raises the similarity of segmented image features, has stronger resistance to noise interference than other algorithms, and significantly improves the recognition accuracy of different parts of the image. The improved algorithm provides a reference for subsequent image processing research.