公开(公告)号：WO2018215907A1

公开(公告)日：2018-11-29

申请号：PCT/IB2018/053555

申请日：2018-05-21

Applicant: NAHATA, Apul ,  GARG, Ashwani ,  KURUVILLA, John

Inventor： NAHATA, Apul ,  GARG, Ashwani ,  KURUVILLA, John

IPC: G06T7/00 ,  G06K9/68

CPC classification number: G06K9/00744 ,  G06T7/20 ,  G06T2207/10016 ,  G06T2207/20016

Abstract: The present disclosure relates to a method for enabling accurate object detection in a video having multiple image frames, wherein the method comprises, for at least one of the multiple image frames; sampling down the at least one image frame to generate multiple temporary image frames of lower resolution; performing pixel level object detection on temporary image frame having lowest resolution to generate a corresponding temporary binary image that comprises a first binary value indicating the object being detected and a second binary value indicating the object not being detected; and iteratively interpolating the temporary binary image to next higher resolution, and refining the interpolated temporary binary image based on a corresponding temporary image frame having same higher resolution till the next higher resolution is equal to that of the at least one image frame, wherein the interpolated temporary binary image is representative of accurately detected object.