Ekualisasi Histogram Dan Algoritma Kultural Untuk Segmentasi Citra Pantai

Abstract

This paper proposes a coastal image segmentation framework to cluster land and ocean pixel intensities to obtain coastline. The segmentation method commonly used is segmentation with the Otsu and Kapur methods, the problem with this method is that it can only segment images with one threshold value, and the complexity increases as the complexity of the image being processed increases. So that it produces output with less than optimal results and requires a long processing time. Therefore, an optimization method is proposed using histogram equalization which aims to increase the contrast value of the test image so that it is easier to segment, then proceed with image clustering using a cultural algorithm with multiple threshold values, so that the intensity of pixels representing land and water areas can be clustered with good. The target of the application of these two methods is to optimize in order to improve the performance of the test image segmentation. In addition, the advantage of this clustering method is that it produces a compressed output image so that it will be more efficient in terms of further processing, storage, and data transmission. The measurement methods applied are the measurement of processing time, Feature Similarity Index Measure (FSIM), and Structural Similarity Index Measure (SSIM) to compare the similarity of features and structures between the test image and the resulting output image. Where the proposed method has been able to cluster classes based on the configured threshold value, so as to be able to segment land and water areas.