中科微点-全球专利检索

  1. Login
  2. Sign Up

Search Results

50 records (took 0.022 seconds)

    TRAINING SYSTEM, TRAINING METHOD AND RECOGNITION SYSTEM
    1.
    发明公开
    TRAINING SYSTEM, TRAINING METHOD AND RECOGNITION SYSTEM 审中-公开

    公开(公告)号：US20240273944A1

    公开(公告)日：2024-08-15

    申请号：US18236142

    申请日：2023-08-21

    Applicant: REALTEK SEMICONDUCTOR CORP.

    Inventor： Chien-Hao Chen ,  Chih-Wei Wu ,  Shih-Tse Chen

    IPC: G06V40/16 ,  G06V10/82

    CPC classification number: G06V40/172 ,  G06V10/82

    Abstract: A training system, a training method, and a recognition system are provided. The training method is used to train a neural network module including: an encoder module, a shared decoder module, a synthesis module, and a classification module. The training method includes performing in a training epoch: repeatedly executing: taking a training image from a training set as an input image, obtaining a first loss based on training feature images of the training image and the feature images corresponding to the training image, and obtaining a second loss based on a classification marker of the training image and a classification generated by the classification module in correspondence with the training image; and updating first parameters and second parameters based on an average value of all the first losses and an average value of all the second losses obtained in the preceding step and an update algorithm.

    COMPRESSION METHOD AND ASSOCIATED ELECTRONIC DEVICE
    2.
    发明公开
    COMPRESSION METHOD AND ASSOCIATED ELECTRONIC DEVICE 审中-公开

    公开(公告)号：US20230206389A1

    公开(公告)日：2023-06-29

    申请号：US18088794

    申请日：2022-12-27

    Applicant: Realtek Semiconductor Corp.

    Inventor： Tsung-Hsuan Li ,  Wan-Ju Tang ,  Shih-Tse Chen

    IPC: G06T3/40 ,  G06T9/00

    CPC classification number: G06T3/40 ,  G06T9/00

    Abstract: The present invention provides a compression method, wherein the compression method includes the steps of: setting a quantization function; setting a plurality of scaling ratios; receiving image data; for a block of a frame of the image data, using a conversion matrix to convert data of the block to generate a plurality of converted data; determining a specific scaling ratio from the plurality of scaling ratios according to the plurality of converted data; and using the quantization function to perform a quantization operation on a plurality of adjusted data to generate compressed data, wherein the plurality of adjusted data are generated according to the plurality of converted data and the specific scaling ratio.

    Motion image integration method and motion image integration system capable of merging motion object images
    4.
    发明授权
    Motion image integration method and motion image integration system capable of merging motion object images 有权

    公开(公告)号：US11270442B2

    公开(公告)日：2022-03-08

    申请号：US16745258

    申请日：2020-01-16

    Applicant: Realtek Semiconductor Corp.

    Inventor： Chao-Hsun Yang ,  Shang-Lun Chan ,  Chun-Chang Wu ,  Shih-Tse Chen

    IPC: G06T7/215 ,  G06T7/246

    Abstract: A motion image integration method includes acquiring a raw image, detecting a first motion region image and a second motion region image by using a motion detector according to the raw image, merging the first motion region image with the second motion region image for generating a motion object image according to a relative position between the first motion region image and the second motion region image, and cropping the raw image to generate a sub-image corresponding to the motion object image according to the motion object image. A range of the motion object image is greater than or equal to a total range of the first motion region image and the second motion region image. Shapes of the first motion region image, the second motion region image, and the motion object image are polygonal shapes.

    Method for generating target gain value of wide dynamic range operation
    6.
    发明授权
    Method for generating target gain value of wide dynamic range operation 有权

    公开(公告)号：US09800793B2

    公开(公告)日：2017-10-24

    申请号：US15296324

    申请日：2016-10-18

    Applicant: Realtek Semiconductor Corp.

    Inventor： Wei-Te Chang ,  Shih-Tse Chen

    IPC: H04N5/235 ,  G06T7/40 ,  G06K9/46 ,  G06K9/62

    CPC classification number: H04N5/2352 ,  G06K9/4661 ,  G06K9/6215 ,  G06T7/408 ,  G06T7/90

    Abstract: A method for generating a target gain value of a wide dynamic range (WDR) operation is disclosed including: acquiring an average bright portion luminance corresponding to an average pixel luminance of a bright portion of a video image; acquiring an average dark portion luminance corresponding to an average pixel luminance of a dark portion of the video image; generating an initial gain value of the WDR operation according to a difference between the average bright portion luminance and the average dark portion luminance; and adjusting the initial gain value according to at least one of a color temperature and an exposure duration configuration value of the video image to generate the target gain value.

    Image correction method using approximately non-linear regression approach and related image correction circuit
    8.
    发明授权
    Image correction method using approximately non-linear regression approach and related image correction circuit 有权
    使用近似非线性回归方法和相关图像校正电路的图像校正方法

    公开(公告)号：US09269130B2

    公开(公告)日：2016-02-23

    申请号：US13858101

    申请日：2013-04-08

    Applicant: Realtek Semiconductor Corp.

    Inventor： Shih-Tse Chen ,  Hao-Tien Chiang

    IPC: G06T5/00

    CPC classification number: G06T5/006

    Abstract: An image correction method arranged for processing an original image to obtain a corrected image includes steps: receiving the original image from an image sensor; regarding each pixel of the original image, calculating a horizontal distance and a vertical distance between the pixel and a reference point in the original image; determining a horizontal ratio parameter and a vertical ratio parameter according to the horizontal distance and the vertical distance between the pixel and the reference point in the original image; and performing an approximately non-linear regression calculation on the horizontal ratio parameter, the vertical ratio parameter and a coordinate of the pixel to obtain a position of the pixel in the corrected image.

    Abstract translation: 一种布置用于处理原始图像以获得校正图像的图像校正方法，包括以下步骤：从图像传感器接收原始图像; 关于原始图像的每个像素，计算原始图像中的像素和参考点之间的水平距离和垂直距离; 根据原始图像中的像素与参考点之间的水平距离和垂直距离确定水平比参数和垂直比率参数; 并且对水平比参数，垂直比参数和像素的坐标执行近似非线性回归计算，以获得校正图像中的像素的位置。

    NEURAL NETWORK SYSTEM AND SIGNAL PROCESSING METHOD
    9.
    发明公开
    NEURAL NETWORK SYSTEM AND SIGNAL PROCESSING METHOD 审中-公开

    公开(公告)号：US20240273887A1

    公开(公告)日：2024-08-15

    申请号：US18216147

    申请日：2023-06-29

    Applicant: REALTEK SEMICONDUCTOR CORP.

    Inventor： Chih-Wei Wu ,  Chien-Hao Chen ,  Wei-Hsiang Shen ,  Shih-Tse Chen

    IPC: G06V10/82 ,  G06V10/774

    CPC classification number: G06V10/82 ,  G06V10/774

    Abstract: A neural network system and a signal processing method are provided. The neural network system includes at least one processing unit and a neural network module. The signal processing method includes: inputting a neural network input to the neural network module by the processing unit to generate an input at a previous layer of each convolutional transformer layer; performing pointwise convolution on the input by a key embedding layer based on key convolutional kernels to output a key tensor; performing convolution on the input by a value embedding layer based on value convolutional kernels to output a value tensor; performing a convolution on the cascading tensor of a first tensor and the key tensor by an attention embedding layer based on attention convolution kernels to output an attention tensor; and outputting an output tensor based on the attention tensor and the value tensor by an output module.

    DETECTION CIRCUIT AND ASSOCIATED DETECTION METHOD
    10.
    发明公开
    DETECTION CIRCUIT AND ASSOCIATED DETECTION METHOD 审中-公开

    公开(公告)号：US20240257370A1

    公开(公告)日：2024-08-01

    申请号：US18232364

    申请日：2023-08-10

    Applicant: Realtek Semiconductor Corp.

    Inventor： Chih-Yuan Koh ,  Shih-Tse Chen

    IPC: G06T7/50 ,  G06T7/70

    CPC classification number: G06T7/50 ,  G06T7/70 ,  G06T2207/20081 ,  G06T2207/20084 ,  G06T2207/30196

    Abstract: The present invention provides a detection circuit including a neural network module and a calculation circuit is disclosed. The neural network module is configured to receive an image to generate an output tensor, wherein the output tensor includes position information of a specific object and distance adjustment information. The calculation circuit is coupled to the neural network module, and is configured to calculate an initial distance between an image capture device and the specific object according to the position information of the specific object, and generate an estimated distance according to the initial distance and the distance adjustment information.

Patent Agency Ranking