Classification of Segmented Images for Analysis Using Hybrid Methodology

Abstract With natural and aerial images, besides the absence of exact damage, images are of a relatively poor quality and then analysis in general is complex due to data composition, described in terms of speckle formation. Due to discolor creation and reduction objects, it is complicated to correctly divide the ultrasound image to identify concerned objects with the exact location and shape. There are a huge number of diverse approaches are used on segmenting an images freshly engaged. Seeded region growing method is mostly applied for image segmentation based on region. But this method fails to process since the region at the edge of the image is not processed well. To overcome this issue, the previous work presented Multitude Regional Texture Extraction for Image Segmentation of Aerial and Natural Images. Based on multitude region, extraction of texture is to be done for image segmentation. In this work, we plan to present a hybrid methodology for classification of segmented images in an efficient manner. A hybrid unsupervised/supervised classification methodology is applied to aerial and natural images and described briefly. The hybrid method varies from the conservative classification intellect to that the clustering algorithm is useful to a set of regions that are acquired from the segmented image. The geometric parameters of these regions are utilized to categorize based on the results attained. After the prior step, some regions are chosen to be training data sets on a supervised categorization step. An evaluation is done among the pixel per pixel classification and the region classification. The experimental evaluation is conducted with training samples of natural and aerial images to show the performance of the proposed classification of segmented images for analysis using hybrid methodology and compare the results with an existing seeded region growing model and Multitude Regional Texture Extraction for Image Segmentation.