On Regenerative and Discriminative Learning from Digital Heritages: A Fractal Dimension Based Approach

Document Type

Conference Article

Publication Title

Lecture Notes in Networks and Systems

Abstract

In recent years, analyzing historical monuments using computer vision techniques has become extremely important for renovation, tourism experience, and preservation. While such studies are common for European architectural research, their findings are not directly applicable to Indian Monuments as the latter has significant differences in architecture and ornamental design. A key feature of Indian religious monuments such as temples is the heavily ornamented spires containing complex geometrical motifs and sculptures. However, a lack of available datasets prevents a detailed analysis of spires found in Indian monuments. Fractal Dimension can act as a measure of the architectural complexities by engaging the curvature present in the image of a monument. In this paper, we exploit the notion of Fractal Dimension through patch-based image analysis to visualize the architectural complexity of Indian monuments. Further, we present a technique to segment the original image such that the areas with rich architectural patterns stays similar while the regions with no marks or designs change to zero. The segmented image from the initial and fractal-based reconstruction gives a substantial recall improvement to a classifier compared to the original, proving the claim of using the rich segmentation mask from the fractal dimension-based reconstruction.

First Page

405

Last Page

417

DOI

10.1007/978-981-99-3250-4_31

Publication Date

1-1-2023

This document is currently not available here.

Share

COinS