<![CDATA[Traditional Minangkabau ornaments such as pucuak rabuang, itiak pulang patang, kaluak paku, and rabuang sanjo represent a form of visual cultural heritage with high aesthetic and philosophical value. However, the digital documentation and preservation of these ornaments still face significant challenges, particularly due to variations in media, fine surface textures, uneven illumination, and complex image backgrounds. These conditions complicate the separation of ornament motifs from the background and consequently affect the accuracy of identification and classification processes. This study aims to develop an image processing approach for the detection and identification of Minangkabau ornaments through image quality enhancement and multi-level segmentation. The proposed method begins with a preprocessing stage that includes motif area cropping, image size normalization, noise reduction through filtering, contrast stretching, and Contrast Limited Adaptive Histogram Equalization (CLAHE) to enhance local contrast. Subsequently, segmentation is performed using the Multi-Threshold Otsu method to divide the image into multiple intensity classes, enabling a more detailed separation of ornament structures. The segmentation results are evaluated using morphological analysis and further tested using a Convolutional Neural Network (CNN) to assess classification performance. Experiments were conducted on a dataset of 1,024 images, with a training–testing split of 70% and 30%, respectively. The experimental results demonstrate that the proposed approach produces representative motif segmentation and achieves a classification accuracy of 99.67%. These findings indicate that the integration of systematic preprocessing, multi-threshold segmentation, and CNN-based classification is effective in supporting the digital preservation of Minangkabau ornaments.]]>
Copyrights © 2026
