An Approach to Noisy Synthetic Color Image Segmentation Using Unsupervised Competitive Self-Organizing Map

Abstract

The complex data is transformed as the simple but meaningful smaller data groups in the segmentation process. It is the utmost phase of the data exploration process. It is the method of allocating a tag to each pixel to make them as groups (clusters) and the pixels using the same tag have the common characteristics such as color, texture, or intensity. It is challenging to decide on the optimal segmentation method. For noisy images, segmentation becomes more difficult one. This is due to both the image and noisy pixels are considered as the same category. In this work, an artificial neural network based unsupervised self-organizing maps utilized to analyze and cluster the noisy synthetic images. The projected technique employed three levels (competition, cooperation, and adaptation) of competitive learning to segment the data into meaningful regions. The investigational end result undoubtedly revealed the proficiency of the suggested methodology to cluster the noisy images.