Universal Image Distance & Support Vector Machine based Optimized Image Classifier System

In image processing and computer engineering, Image Classification of distantly sensed images is one of the quickly increasing areas of research. Image classification techniques are mainly used for quality control in production engineering, in-depth study of web technologies, medical diagnosis and in several other disciplines. Image processing techniques have broadly used in this application to detect high quality and poor quality area or to type of contamination in a microscopic image and other similar decision making tasks. Other Algorithms like as thresholding, blob analysis, and edge detection, for example, it can be found in every machine vision software vendor's toolbox since these can be used in several applications to solve a quite large number of imaging tasks. Classifications of images are still computationally normal subject, and it is hard to implement using some machine learning approaches, using Neural Networks or Bayesian Classifiers. In recent years, Support Vector Machines is also used for this purpose. Each of these methods follows supervised learning in which the system gives numerous examples of images that are manually labeled. On the contrast, in unsupervised learning, also called clustering, is the approach in which no training data send to the machine than the machine itself has to come up in with the grouping of the input data in the form of clusters. The result data of the previous process, a trained model is processed which is used to predict the features of unknown images. Such traditional supervised learning techniques can use either generative or discriminative methods to perform this task. In this dissertation, UID (Universal Image Distance) techniques are used in an optimized manner to represent an image in the form of a vector quantity. The distance between this representation and that of a prototype image is computed to find the similarity score between the images. This resulting score can be used to train any machine learning system in a supervised or unsupervised environment. In this dissertation, an SVM-based classifier is trained using feature vectors to train in a supervised manner. The precision and accuracy of the machine are calculated over the benchmark techniques of image classification. The overall performances of the proposed methods are evaluated using MATLAB, recall and kappa measure. The result of simulation has efficiency in approach and gives a valid result.