内容紹介

An ordinary camera captures the real 3D scene as a projected 2D image with three channels. During this projection process, some scene information gets lost such as depth and surface normal, and some information gets compressed such as resolution and dynamic range. This presentation answers the questions what information has been lost and how to get them back by introducing the full-dimensional imaging framework. The lost and compressed information is restored as additional channels of a generalized image, e.g., a high-quality depth map with as high resolution as the 2D image, with state of the art computer vision algorithms, computational photography techniques, and novel camera prototyping. The concept of camera intelligence is then introduced to inspire the future research that complements artificial intelligence with camera design for solving visual computing problems that can never be solved using a conventional camera.

詳細情報

刊年

2016

形態

電子化映像資料(1時間15分20秒)

シリーズ名

情報科学研究科・ゼミナール講演 ; 平成28年度

注記

講演者所属: 産業技術総合研究所

講演日: 平成28年10月31日

講演場所: 情報科学研究科大講義室L1

標題言語

英語 (eng)

本文言語

英語 (eng)

著者情報

Shi, Boxin