首页> 外国专利> AI AI AI encoding apparatus and operating method for the same and AI decoding apparatus and operating method for the same

AI AI AI encoding apparatus and operating method for the same and AI decoding apparatus and operating method for the same

机译:AI AI AI编码设备和操作方法的相同和AI解码设备和操作方法

摘要

A memory for storing one or more instructions and a processor for executing one or more instructions stored in the memory, wherein the processor is an image corresponding to a first image in which an AI encoding apparatus AI downscales an original image using a first DNN. When data is obtained, a second image corresponding to the first image is reconstructed based on the image data, and a second image is input to a second DNN including a plurality of layers, a first filter kernel among the plurality of layers and, from the first layer including the second filter kernel, obtains first result values based on the operation result of the second image and the first filter kernel, and is based on the operation result between the second image and the second filter kernel. obtain second result values based on the normalization by converting the first result values to first values based on the first scale factor, and convert the second result values to second values based on the second scale factor by doing normalization, converting the normalized values into integer values included in a preset range, and inputting the converted integer values to a second layer that is the next layer of the first layer among the plurality of layers, and the second layer obtains third result values in , and based on the third result values, a second image acquires an AI-upscaled third image, the second DNN corresponding to the first DNN, a first scale factor and An apparatus for decoding an image in which the second scale factor is determined by values associated with scale factors included in the first DNN is disclosed.
机译:用于存储一个或多个指令和处理器的存储器,用于执行存储在存储器中的一个或多个指令,其中处理器是对应于第一图像的图像,其中AI编码装置AI使用第一DNN缩小原始图像。当获得数据时,基于图像数据重建对应于第一图像的第二图像,第二图像被输入到包括多个层的第二DNN,多个层中的第一滤波器内核,并且来自包括第二滤波器内核的第一层基于第二图像和第一滤波器内核的操作结果获得第一结果值,并且基于第二图像和第二滤波器内核之间的操作结果。通过基于第一刻度因子将第一结果值转换为第一个值来获得第二结果值,并通过执行归一化将第二结果值转换为基于第二比例因子的第二值,将归一化值转换为整数值包括在预设范围内,并将转换的整数值输入到多个层中的第一层的下一层,第二层获得第三结果值,并基于第三结果值,a第二图像获取AI-Uppaled第三图像,对应于第一DNN的第二DNN,第一比例因子和用于解码的图像,其中通过与第一个DNN中包括的比例因子相关联的值来确定第二比例因子的图像是披露。

著录项

  • 公开/公告号KR102312338B1

    专利类型

  • 公开/公告日2021-10-14

    原文格式PDF

  • 申请/专利权人 삼성전자주식회사;

    申请/专利号KR20190078344

  • 发明设计人 딘 쿠오칸;최민석;최광표;

    申请日2019-06-28

  • 分类号H04N19/132;G06N3/08;G06T1/20;G06T1/60;H04N19/184;H04N19/50;H04N19/85;

  • 国家 KR

  • 入库时间 2022-08-24 21:42:35

相似文献

  • 专利
  • 外文文献
  • 中文文献
获取专利

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号