標題: without full decompression 59 [打印本頁] 作者: tsxfoquv 時間: 2016-12-26 18:23 標題: without full decompression 59
Abstract This paper presents a technique for efficient and generic corresponding of compressed video injections, through compact signatures extracted straight without decompression. The compact bank is based on the Dominant Colouring Profile (DCP); a sequence of dominant colours extracted and arranged as a sequence of huge amounts, in analogy to the our retinal representation of a scene. The actual proposed signature represents the video shot with 490 integer Belstaff Coat ideals, facilitating for real time running to retrieve a maximum list of matching videos. The technique may work directly on MPEG compressed movies, without full decompression, as it is utilizing the DC image as a platform for extracting colour characteristics. The DC image incorporates a highly reduced size, whilst retaining most of visual factors, and provides high performance compared to the full I frame. The studies and results on numerous standard datasets show the promising performance, both the accuracy as well as the efficient computation complexity, of your proposed technique.
