Automatic Test Methods for Image and Video Verification

Abstract: In this thesis four methods for automatic verification of images and video on mobile platforms are developed. Both the case of recording images and video and the case of viewing images and video on the mobile lcd screen are considered. The first method is used to test the zoom function of the camera. It uses SURF decriptors along with clustering and histograms to determine which of six discrete zoom levels the current frame belongs to. The second method identifies color effects and color anomalies using histograms. The third method determines if the autofocus works correctly by measuring the average length of edges in the image. The fourth method is an artifact detection scheme using a non-reference implementation of the SSIM metric, used in conjunction with a for this purpose specially designed test setup. Together these methods form a tool kit for detecting the mnost common errors to occur in images and video during the development stage of mobile platforms.