Article: Research on artistic style recognition and image transfer method based on deep visual feature extraction Journal: International Journal of Information and Communication Technology (IJICT) 2025 Vol.26 No.40 pp.86 - 103 Abstract: This research presents a comprehensive approach to artistic style recognition and image style transfer using deep visual feature extraction techniques. To enhance the identification of fine art forms, the study employs a two-stage classification model that combines shallow and deep neural networks, utilising convolutional neural networks (CNNs), namely VGG16 and VGG19. A novel neural style transfer network is proposed, incorporating a coarse-to-fine methodology and whitening and colouring transformation (WCT) to preserve global content structures while effectively applying local stylistic elements. Extensive experiments on the Wiki Art and Pandora 18K datasets validate the model's ability to enhance style classification accuracy, minimise restructuring loss, and reduce runtime. The outcomes show that the suggested approach greatly improves automated art analysis and digital creative apps while keeping high-resolution image integrity and successfully integrating the unique visual traits of artistic styles. Inderscience Publishers - linking academia, business and industry through research

Title: Research on artistic style recognition and image transfer method based on deep visual feature extraction

Authors: Wei Wang; Hanchen Li; Rushana Sulaiman @ Abd Rahim; Peilei Cui

Addresses: College of Creative Arts, University of Malaya, Kuala Lumpur, 50603, Malaya, Malaysia ' College of Creative Arts, University of Technology MARA, Shah Alam, Selangor, 40450, Malaya, Malaysia ' College of Creative Arts, University of Technology MARA, Shah Alam, Selangor, 40450, Malaya, Malaysia ' SophNet, Beijing, 10000, China

Abstract: This research presents a comprehensive approach to artistic style recognition and image style transfer using deep visual feature extraction techniques. To enhance the identification of fine art forms, the study employs a two-stage classification model that combines shallow and deep neural networks, utilising convolutional neural networks (CNNs), namely VGG16 and VGG19. A novel neural style transfer network is proposed, incorporating a coarse-to-fine methodology and whitening and colouring transformation (WCT) to preserve global content structures while effectively applying local stylistic elements. Extensive experiments on the Wiki Art and Pandora 18K datasets validate the model's ability to enhance style classification accuracy, minimise restructuring loss, and reduce runtime. The outcomes show that the suggested approach greatly improves automated art analysis and digital creative apps while keeping high-resolution image integrity and successfully integrating the unique visual traits of artistic styles.

Keywords: artistic style recognition; image style transfer; deep learning; convolutional neural networks; neural style transfer; whitening and colouring transformation; WCT; shallow neural network; SNN; feature extraction; image processing; coarse-to-fine stylisation; computer vision; machine learning; Wiki Art.

DOI: 10.1504/IJICT.2025.149816

International Journal of Information and Communication Technology, 2025 Vol.26 No.40, pp.86 - 103

Received: 02 Aug 2025
Accepted: 01 Sep 2025
Published online: 13 Nov 2025 *

Title: Research on artistic style recognition and image transfer method based on deep visual feature extraction

Keep up-to-date