Modified Satellite Image Segmentation Based on Region Method

Abstract

A modified segmentation technique is proposed to segment images using Univariate statistical parameters such as minimum, maximum, and mean values as well as contour length. This segmentation technique is applied to satellite images of seven bands captured by the Landsat (8, 9) OLI satellite in 2022. The utilization of principal component analysis (PCA) is employed in this study, as it has been observed that the combined value of the initial three principal components encompasses nearly 99% of the pertinent information. To effectively represent the identified elements, the Freeman chain coding technique is applied, ensuring both reduced bounds and low bit rates. The segmentation process for the 1^st PC images involves several stages. This approach employs a threshold value of 0.60 and a minimum contour length of 8. The process commences at the initial stage and advances consistently until it reaches the final stage, ultimately generating an image devoid of any fragmented regions. The suggested segmentation method provides realistic thin and closed borders compared to any method based on the edge-based technique. The results showed that the total number of segment regions with a contour length greater than 6 equals 358 within four categories that have the same characteristic.