Abstract: An image processing method includes obtaining a plurality of frames of raw images. After preprocessing such as image alignment, channel splitting or pixel rearrangement is performed on the obtained plurality of frames of raw images, detail restoration is performed on an image based on a deep learning network, and luminance enhancement and color enhancement are performed on an image output by the deep learning network. A plurality of types of processing related to detail restoration are integrated into a same deep learning network.

Abstract: A video processing method is provided. The method includes: In a video recording process, an electronic device obtains raw image data and metadata, and processes the raw image data by using an image signal processor, to obtain first video data; and when exiting video recording, obtains a first video file, and stores the first video file and the metadata, where the first video file is a video file generated by encoding the first video data. A gallery application invokes required metadata based on a video algorithm that needs to be performed, and processes the first video file to obtain a second video file, where video quality of the second video file is higher than video quality of the first video file.

Abstract: The present disclosure relates to the field of image processing technologies, and discloses a photographing method and apparatus for intelligent framing recommendation, to automatically match an appropriate framing recommendation solution for a user according to an algorithm, so that imaging quality is improved. The method includes: An electronic device obtains at least one framing recommendation result based on images captured by at least two cameras with different focal lengths, where each of the at least one framing recommendation result includes a framing recommendation frame, and the framing recommendation frame indicates a framing recommendation effect of image photographing; and displays a target image based on a framing recommendation result selected by a user among the at least one framing recommendation result, where the target image is obtained through cropping, based on the selected framing recommendation result, the images captured by the at least two cameras.

Abstract: The present disclosure relates to a method and system for reconstructing an image of a midpalatal suture based on a Sobel operator and belongs to the technical field of image processing. Firstly, data preprocessing is performed on a cone beam computed tomography (CBCT) file of a midpalatal suture of a maxilla to obtain a plurality of local images of the midpalatal suture. Merging fusion is then performed on the plurality of local images of the midpalatal suture to obtain a reconstructed image of the midpalatal suture, and a Sobel operator is utilized for optimization during the merging fusion. Thus, an overall profile of a region shape of the midpalatal suture can be reconstructed in a same image from midpalatal suture regions distributed in multiple layers of CBCT images, facilitating intuitive and accurate judgment on the midpalatal suture region by a doctor.

Abstract: This application relates to the field of image processing technologies, and provides a photographing method and apparatus for intelligent framing recommendation, to automatically match an appropriate framing recommendation solution for a user according to an algorithm, so that imaging quality is improved. The method includes: An electronic device obtains at least one framing recommendation result based on images captured by at least two cameras with different focal lengths, where the framing recommendation result includes a framing recommendation frame, and the framing recommendation frame indicates a framing recommendation effect of image photographing; and displays a target image based on a framing recommendation result selected by a user, where the target image is an image obtained through cropping, based on the framing recommendation result, the images captured by the at least two cameras.

Abstract: An image processing method includes obtaining a plurality of frames of raw images. After preprocessing such as image alignment, channel splitting or pixel rearrangement is performed on the obtained plurality of frames of raw images, detail restoration is performed on an image based on a deep learning network, and luminance enhancement and color enhancement are performed on an image output by the deep learning network. A plurality of types of processing related to detail restoration are integrated into a same deep learning network.

Abstract: Type completeness detection methods and systems are provided to determine a browse node completeness score for browse nodes in an electronic catalog. For example, the type completeness detection system may select a browse node, identify a plurality of search queries previously submitted by users of the electronic catalog, determine a plurality of items that are frequently selected after submitting the search queries, calculate a query completeness score for each search query, determine a browse node completeness score for the browse node, and initiate an action in response to determining that the browse node completeness score falls below a threshold value.

Abstract: A hard-decision (HD) forward error correcting (FEC) coded signal is decoded by a decoder to produce decoded bits using marked reliable bits of the HD-FEC coded signal and marked unreliable bits of the HD-FEC coded signal. The marked reliable and unreliable bits are computed by calculation and marking blocks based on an absolute value of log-likelihood ratios of the HD-FEC coded signal. The HD-FEC coded signal may be, for example, a staircase code coded signal or a product code coded signal.

Abstract: A hard-decision (HD) forward error correcting (FEC) coded signal is decoded by a soft-aided decoder (112) to produce decoded bits (118) using marked reliable bits of the HD-FEC coded signal and marked unreliable bits of the HD-FEC coded signal (soft-aided bit marking) that are computed by calculation (114) and marking blocks (116) based on an absolute value of log-likelihood ratios (LLRs) of the HD-FEC coded signal. The hard-decision (HD) forward error correcting (FEC) coded signal may be, for example, a staircase code (SCC) coded signal or a product code (PC) coded signal.

Abstract: A time information coding method is provided, to solve the problem of coded values low calculation efficiency, which is resulted by adopting the existing time information coding schemes. The method comprises: determining time information to be coded; coding the time information to be coded to a first integer with a specified number of bits under a first time scale; coding the first integer into a second integer with a specified number of bits under a second time scale, the second integer being as a coded value under the second time scale of the time information to be coded. A time information coding device, a searching method for coded values, a decoding method and device are also provided.

Abstract: A panel light apparatus is designed to be installed to a ceiling. The panel light apparatus has a bracket firstly fixed to the ceiling. The bracket has a first lateral side and a second lateral side. The panel light apparatus has a detachable panel module to be attached to the bracket. The detachable panel module has a panel, a surrounding wall, a back cover, a first connector and a second connector. When the detachable panel module is attached to the bracket, the first connector and the second connector respectively have insertion portions kept in holes of the first lateral side and the second lateral side of the bracket.