Contour Feature Extraction of Medical Image Based on Multi-Threshold Optimization

Abstract

During the process of fine segmentation of medical images, although a single threshold can improve the efficiency of processing, there will be the problem of fuzzy features and non-convergence of threshold in denoising of details such as contour extraction. To extract contour information of medical images, a method based on multi-threshold optimization is proposed. This paper analyzes the influence of contour wave transformation on gray correlation degree and noise intensity of different medical images and improves the Bayesian threshold. The middle threshold function was improved by correlation characteristics of contour wave coefficients, and contour features of medical images were constrained by multiple thresholds. Based on the above, the dimension of the medical image was reduced by the wavelet multi-resolution analysis method, and the corresponding threshold search space was obtained. A genetic algorithm was used to find the best quasi threshold in the search space. Through this value, the attribute histogram of the medical image was established, the best feature extraction threshold of the medical image was obtained by the golden section method, and contour feature information of the medical image was extracted. The experimental results show that the proposed method can achieve the fast extraction of the contour feature information of running image, get an ideal feature extraction effect, and has high efficiency of feature extraction.