Summarization of Wireless Capsule Endoscopy Video Using Deep Feature Matching and Motion Analysis

Conventional Wireless capsule endoscopy (WCE) video summary generation techniques apprehend an image by extracting hand crafted features, which are not essentially sufficient to encapsulate the semantic similarity of endoscopic images. Use supervised methods for extraction deep features from need enormous amount accurate labelled data training process. To solve this, we use unsupervised learning method extract using convolutional auto encoder. Furthermore, WCE images classified into similar and dissimilar pairs fixed threshold derived through large number experiments. Finally, keyframe based on motion analysis is used derive a structured video. Proposed achieves average F-measure 91.1% with compression ratio 83.12%. The results indicate that proposed more efficient compared existing summarization techniques.