MULTITHRESHOLDING IN GRAYSCALE IMAGE USING PEA FINDING APPROACH AND HIERARCHICAL CLUSTER ANALYSIS

Abstract Image segmentation is typically used to distinguish objects that exist in an image. However, it remains difficult to accommodate favourable thresholding in multimodal image histogram problem with specifically desired number of thresholds. This research proposes a novel approach to find thresholds in multimodal grayscale image histogram. This method consists of histogram smoothing, identification of peak(s) and valley(s), and merging process using hierarchical cluster analysis. Using five images that consisted of grayscale and converted-to-grayscale images. This method yields maximum value of accuracy, precision, and recall of 99.93%, 99.75%, and 99.75% respectively. These results are better than the similar peak finding method in multimodal grayscale image segmentation.