公开(公告)号：US09105088B1

公开(公告)日：2015-08-11

申请号：US14046919

申请日：2013-10-04

Applicant: Google Inc.

Inventor： Ryan Petrie ,  Thor Carpenter

IPC: G06K9/40 ,  G06T5/00

CPC classification number: G06T5/003 ,  G06T5/002 ,  G06T5/004 ,  G06T5/50 ,  G06T2207/10004 ,  G06T2207/20012 ,  G06T2207/20192 ,  G06T2207/20221 ,  G06T2207/30201

Abstract: Implementations relate to blurring an image while preserving details in the image. In some implementations, a method includes receiving an input image and creating a blurred image from the input image by averaging pixel intensities of a plurality of pixels in the input image. The method creates a transparency mask from the input image, where the transparency mask indicates transparency values for pixel positions of the input image based on edges detected in the input image. The pixel intensities of the input image and pixel intensities of the blurred image are composited based on the transparency values of the transparency mask to create an output image having greater blur than the input image in areas outside the detected edges.

Abstract translation: 实现涉及模糊图像同时保留图像中的细节。 在一些实现中，一种方法包括通过对输入图像中的多个像素的像素强度进行平均来从输入图像接收输入图像并创建模糊图像。 该方法从输入图像创建透明度掩码，其中透明掩码基于输入图像中检测到的边缘来指示输入图像的像素位置的透明度值。 基于透明度掩模的透明度值来合成输入图像的像素强度和模糊图像的像素强度，以在检测到的边缘之外的区域中产生具有比输入图像更大的模糊的输出图像。

公开(公告)号：US09060095B2

公开(公告)日：2015-06-16

申请号：US13802241

申请日：2013-03-13

Applicant: GOOGLE INC.

Inventor： Frank Petterson ,  Janahan Vivekanandan ,  Thor Carpenter

IPC: H04N7/14 ,  H04N7/15 ,  G06K9/00

CPC classification number: H04N7/157 ,  G06K9/00228

Abstract: Implementations generally relate to modifying an appearance of a participant during a video conference. In some implementations, a method includes obtaining at least one frame from a media stream, where the at least one frame includes a first face, and where the first face is an image of a face associated with a user. The method also includes determining a plurality of coordinates within the at least one frame. The method also includes obtaining at least one media content item. The method also includes adding the at least one media content item to the at least one frame based on the plurality of coordinates.

Abstract translation: 实现通常涉及在视频会议期间修改参与者的外观。 在一些实施方式中，一种方法包括从媒体流获得至少一个帧，其中所述至少一个帧包括第一面，并且其中所述第一面是与用户相关联的面部的图像。 该方法还包括确定至少一个帧内的多个坐标。 该方法还包括获得至少一个媒体内容项目。 该方法还包括基于多个坐标将至少一个媒体内容项添加到至少一个帧。