Although there are several devices for screening the digestive tract, Wireless Capsule Endoscopy (WCE) is now the gold standard for a non-invasive viewing of small intestine. However, in each examination, more than 55,000 images are recorded. Reading so many images can be time consuming. Thus, image processing and vision recognition techniques are being created to help doctors to save the time for diagnosis. This paper presents a comprehensive overview of WCE devices and recent progresses in vision-based techniques on reducing the time needed for WCE video reading. The recently developed techniques can be classified into three categories: disease (bleeding) detection, image-level and video-level summarizations. Firstly, the features and classification strategies employed for automatic bleeding, disease or abnormal detection are presented. Second, this paper presents the image-level summarization based on motion estimation, C-mean clustering and epitome techniques. Third, we present and compare the various color-bar based global-level WCE video summarizations. The capabilities and advantages of different techniques for improving the efficiency of diagnosis are evaluated.