Wataru Oshiumi, Ankur Dixit, Hiroaki Wagatsuma
Kyushu Institute of Technology, 2-4 Hibikino, Wakamatsu, Kitakyushu, 808-0196, Japan
pp. 163–168
ABSTRACT
A refinement process by human experts is still needed for areas/feature extractions of interest from aerial images in multiple map makers. The construction/road edge extraction is one of them which is highly important in the preprocessing stage for map-making, while it is difficult to isolate man-made structures from a natural landscape involving different size objects. In the present study, we focused on an actual procedure in the decomposition of the Morphological Component Analysis known as MCA to extract specific patterns as serial and parallel models. In our computer experiments, dictionaries of Curvelet and Local Discrete Cosine Transform known as LDCT were introduced for the MCA decomposition and then two models demonstrated a non-negligible difference in the feature extraction performance. This result may contribute to the extension of future possibilities of structural data analyses especially for buildings and roads from shapes in nature.

ARTICLE INFO
Article History
Received 25 November 2022
Accepted 31 March 2023

Keywords
Image processing
Edge detection
MCA
CurveletLocal discrete transform

JAALR3307

Download article(PDF)