The Laplace transform algorithm serves as a pivotal component in the application of technology to humanities. This paper integrates the Laplace transform algorithm with a multi-scale feature learning network to extract digital image features, leveraging the unique value of cultural heritage. By utilizing the discrete convolution kernel structure of the Laplace operator, it performs second-order differential calculations on image edge features. Combined with the encoding and decoding steps of the multiscale feature learning network, this approach enhances image segmentation accuracy. It achieves edge structure extraction and global semantic integration in digital images. Compared with similar image feature extraction methods, the proposed method achieves over 90% accuracy in feature extraction and classification performance across three datasets. Moreover, the computational time ranges from 36.53 to 75.42 seconds. This method demonstrates high precision and fast speed in image feature extraction.
