Video processing architecture: a solution for endoscopic procedures results

Abstract

In this paper we propose an architecture for processing endoscopic procedures results. The goal is to create a complete system capable of processing any type of endoscopic multimedia results, in order to overcome the most common issues in the endoscopic domain (e.g. video’s long-duration, gastroenterologist's possible difficulty to maintain the focus and efficiency during the viewing process, imperfections in images/videos). It was this scenario that led to the conception of the MIVprocessing solution, which will address these and other problems, providing an added value to the elaboration of diagnoses. The MIVprocessing is composed of five tasks: Video Summarization (elimination of the “non-informative” frames); Pre-Processing (correction/improvement of the frames); Pre-Detection; Segmentation; and Feature Extraction and Classification. The idea is to create a framework that brings together the capabilities of different but at the same time complementary concepts (e.g. image and signal processing, machine learning, computer vision). This conjugation applied to the endoscopic domain provides a set of features capable of improving the gastroenterologist's activities during and after the procedure.