Image segments visualizer

Abstract

We need to segment images so that it would become easier for the computer to understand image data. Image segmentation is the process of differentiating heterogeneous parts of the image or group similar neighboring pixels together, that could be used for the further applications such as: Image understanding model, Robotics, Image analysis, Medical diagnosis, etc. The people working on image segmentation face a lot of problems in finding best segmentation algorithm according to their image dataset. Because segmentation tools available in market provide limited image segmentation techniques and also comparison of segmentation algorithms is visual not quantitative, these tools are not free and also mostly tools are online, dependent on internet connection so we are providing a platform where mostly segmentation algorithms are implemented and comparison of these algorithm results with groundtruth are visual as well as quantitative. Image Segments Visualizer(ISV) is a platform which provide implementation of multiple image segmentation algorithms includes (Otsuthresh, Graythresh, Multithresh, Adaptivethresh, Grayconnected, Canny, Sobel, Prewitt, Log, Active contour, Fast marching, Grabcut, Lazysnapping, Imseggeodestic, Particle Swarm Optimization, Darwinian Particle Swarm Optimization, Fractional Order Darwinian Particle Swarm Optimization, Jaccard distance, Dice coefficient and Boundary F1 Score). Some of these algorithms take image and provide binary mask whereas some represent each segment with different color and some only display foreground. So a person can upload his desired image dataset and analyze the results of different available image segmentation algorithms. ISV also provide comparison of segmented image with groundtruth visually by highlighting un similar regions and also gives percentage of similarity based on Jaccard, dice coefficient, bfscore. ISV has three main modules first one is for the segmentation of greyscale image. Second module allow user to segment colored(RGB) images and finally the third one is for comparing results of segmented image with ground truth so that user can find the quality of segmentation algorithm. To implement this project matlab is used which provide us all the api’s to build GUI and backend of ISV. This project is implemented in professional way according to rules of software engineering and it is covering mostly available image segmentation algorithms.