Abstract
Conference Title: 2015 6th IEEE International Conference on Software Engineering and Service Science (ICSESS) Conference Start Date: 2015, Sept. 23 Conference End Date: 2015, Sept. 25 Conference Location: Beijing, China Software testing has been a challenging job over the decades and possess more challenges for complex inputs such as images. While evaluating correctness of the output images, there may exist a large number of correct or incorrect images with insignificant differences. A test oracle is required to evaluate the correctness of output images which may not be available in most of the cases. Currently, output images are evaluated by domain experts such as medical experts, which involves manual inspection of output images at each step of software development. In this paper, we have proposed a mechanism to automate the test oracle using support vector machine. It requires a few correct and incorrect images for the training and is capable of classification of correct and incorrect output images. For the demonstration purpose, we used different implementations of image dilation and compared the results with statistical oracle and metamorphic testing. The results in our initial experiments are encouraging.