-
公开(公告)号：EP3099044B1
公开(公告)日：2018-12-26
申请号：EP14891628.1
申请日：2014-05-15
发明人： ZHU, Congchao , ZHENG, Zuquan , LUO, Wei
CPC分类号： H04N5/357 , G06T5/002 , G06T5/50 , G06T2207/10021 , G06T2207/20182 , H04N5/21 , H04N5/341
摘要： Embodiments of the present invention provide a terminal, including: M cameras, configured to continuously shoot a same scene for K times, where M‰¥2, K‰¥2, and each of the M cameras simultaneously collects one frame of image at each time of shooting; and a processor, connected to the M cameras and configured to perform registration on images obtained at the K times of shooting, and perform temporal and spatial filtering on the registered images. In the embodiments of the present invention, multiple cameras are installed on a terminal. Because the multiple cameras simultaneously collect an image at each time of shooting, a time interval at which a particular quantity of images are collected by the terminal is shorter than that at which the particular quantity of images are collected by a single camera, and a probability of an occurrence of hand trembling or object motion within the time interval also correspondingly decreases. Therefore, a synthesized image has better quality in comparison with an image synthesized by multiple frames of images that are collected by the single camera by means of continuous shooting.
-
公开(公告)号：EP4024323A1
公开(公告)日：2022-07-06
申请号：EP20866370.8
申请日：2020-07-22
发明人： WANG, Yinting , ZHANG, Xi , LI, Ruihua , ZHANG, Liping , HUANG, Yining , LEI, Yi , ZHENG, Zuquan , ZHANG, Yifan , LI, Xin , YANG, Xiaowei
摘要： Embodiments of this application disclose an image processing method, a framework, and an apparatus in a computer vision technology in the field of artificial intelligence. The method includes: obtaining a plurality of frames of raw images; and after preprocessing such as image alignment, channel splitting, or pixel rearrangement is performed on the obtained plurality of frames of raw images, performing detail restoration on an image based on a deep learning network, and performing luminance enhancement and color enhancement on an image output by the deep learning network. In embodiments of this application, a plurality of types of processing related to detail restoration are integrated into a same deep learning network. This avoids mutual impact between different processing when a plurality of types of processing are serially performed, effectively fuses effective information of a plurality of frames of images, and better improves an image processing effect.
-
公开(公告)号：EP3099044A1
公开(公告)日：2016-11-30
申请号：EP14891628.1
申请日：2014-05-15
发明人： ZHU, Congchao , ZHENG, Zuquan , LUO, Wei
IPC分类号： H04M1/725
CPC分类号： H04N5/357 , G06T5/002 , G06T5/50 , G06T2207/10021 , G06T2207/20182 , H04N5/21 , H04N5/341
摘要： Embodiments of the present invention provide a terminal, including: M cameras, configured to continuously shoot a same scene for K times, where M≥2, K≥2, and each of the M cameras simultaneously collects one frame of image at each time of shooting; and a processor, connected to the M cameras and configured to perform registration on images obtained at the K times of shooting, and perform temporal and spatial filtering on the registered images. In the embodiments of the present invention, multiple cameras are installed on a terminal. Because the multiple cameras simultaneously collect an image at each time of shooting, a time interval at which a particular quantity of images are collected by the terminal is shorter than that at which the particular quantity of images are collected by a single camera, and a probability of an occurrence of hand trembling or object motion within the time interval also correspondingly decreases. Therefore, a synthesized image has better quality in comparison with an image synthesized by multiple frames of images that are collected by the single camera by means of continuous shooting.
摘要翻译： 本发明的实施例提供了一种终端，包括：M个摄像机，被配置为连续拍摄K次的相同场景，其中M‰¥2，K‰¥2，并且M个摄像机中的每一个同时在每个摄像机上同时收集一帧图像 射击时间 以及处理器，连接到M摄像机，并被配置为对在拍摄的K次获得的图像进行登记，并对登记图像执行时间和空间过滤。 在本发明的实施例中，多个照相机安装在终端上。 由于多个照相机在每次拍摄时都同时收集图像，所以终端收集特定数量的图像的时间间隔比由单个照相机收集特定数量的图像的时间间隔短，并且概率 在时间间隔内手抖或物体运动的发生也相应地减少。 因此，与通过连续拍摄由单个照相机收集的多帧图像合成的图像相比，合成图像具有更好的质量。
-
-
搜索结果
3 条记录 (耗时 0.01 秒)
