-
公开(公告)号：EP0267967B1
公开(公告)日：1993-08-11
申请号：EP87903422.1
申请日：1987-05-23
申请人： FANUC LTD.
IPC分类号： G06F15/66 , G06F15/336 , G06F15/68
CPC分类号： G06T5/20 , G06F17/153
摘要： In carrying out time division processing of the convolution operation with coefficient matrix data on each of the pixel data of a frame memory (3) which stores a plurality of pixel data, the image processing apparatus of the invention delays the result of addition from an adder means (5) by a predetermined period through a delay means (7), permits the result of intermediate operation to be reliably written onto a buffer means (9), and at the same time corrects the read address of the buffer means (9) by an amount by which the write timing is delayed, in order to expedite the time-division convolution processing.
-
22.发明公开Configurable correlation windows for the direct measurement of differential field distortion 失效用于直接测量差异场失真的可配置关联窗口
公开(公告)号：EP0353644A3
公开(公告)日：1991-05-15
申请号：EP89113939.6
申请日：1989-07-28
发明人： Nishihara, H. Keith
IPC分类号： G06F15/70
CPC分类号： G06K9/6203 , G06F17/153 , G06T7/20
摘要： Area correlation of images or regions of images may be used directly to measure field distortion such as rotation, expansion, or shear, once translation disparities are removed between a pair of images. A method is disclosed for allowing relative field distortions to be applied between a pair of image regions as they are correlated. In binocular stereo, the method enables measurement of elevation gradients and surface curvatures. In motion analysis, the method enables the direct measurement of flow field parameters, while in template matching, rotation and magnification may be measured without the necessity of multiple templates for the same object.
-
公开(公告)号：EP0387849A2
公开(公告)日：1990-09-19
申请号：EP90104837.1
申请日：1990-03-14
IPC分类号： G06F15/336 , G06F15/70
CPC分类号： G06T7/20 , G06F17/153 , H04N19/523
摘要： In a correlation computing device including circuits for segmenting (1, 2, 3) a digital input signal into two kinds of digital signals according to a predetermined rule, arithmetically processing (4, 6) the two digital signals for each of the segments, and cumulatively adding (7) the results of signal processing for each of the segments, a high-order bit elimination circuit (9) is provided for eliminating high-order bits of the digital input signal prior to the signal processing.
摘要翻译： 在包括用于根据预定规则将数字输入信号分割为两种数字信号的电路的相关计算装置中，对每个段的两个数字信号进行算术处理（4,6），以及 累加了（7）每个段的信号处理结果，提供了一个高阶位消除电路（9），用于消除信号处理之前数字输入信号的高位。
-
公开(公告)号：EP0386181A1
公开(公告)日：1990-09-12
申请号：EP89907913.0
申请日：1989-07-03
发明人： LO, Thomas, K. , SACKS, Jack, M. , SIMONI, Wayne, P. , LEE, Leonard, A. , NGUYEN, Harrison , KARLIN, Lynne, M. , QUAN, Andy, L. , HOSOKAWA, Kazuo, R.
CPC分类号： G06T7/20 , G06F17/153 , H04N19/43 , H04N19/51
摘要： On a mis au point un procédé et un appareil mettant en corrélation deux signaux représentant une image en direct et une image de référence. Le signal de référence est traité pour fournir un bit de polarité et un bit de masquage pour chaque position de pixel dans l'image de référence. Le signal vidéo traité est stocké dans une mémoire (20) et utilisé dans une partie (18) à circonvolutionneur et additionneur dans laquelle les bits de polarité entre les images de référence et les images de direct sont déclenchés par la porte (17), et sont utilisés pour produire une sortie de corrélation provenant de l'additionneur (23). Toutefois, les bits de polarité n'influent pas sur la sortie de corrélation si la sortie logique provenant de la porte (19) est dans un état donné se produisant si les bits de masquage associés indiquent que la valeur de gradient au niveau de la position particulière de pixel ne dépasse pas certaines valeurs de seuil minimum.
-
公开(公告)号：EP0229919A2
公开(公告)日：1987-07-29
申请号：EP86116024.0
申请日：1986-11-18
CPC分类号： G06F17/153 , H04N1/40056 , H04N1/401
摘要： A flexible image acquisition and processing system comprising photosensor means (15) supplying analogue output signals (20) for compensating and digital conversion means (21) for supplying digital output signals (22) representing the image; timing means (27) for controlling scanning of the photosensor means (15); means (105) for performing two-dimensional convolutions of data representing the images; and system operating phase control means (13). The main characteristic of the aforementioned system consists in the fact that the control means (13) comprise a processing unit for calculating first compensating values for standardising to a given value the analogue output signals (20) for single photoelements on the aforementioned photosensor means (15) under dark conditions, and for calculating second compensating values for standardising the output signals for the aforementioned single photoelements on the aforementioned photosensor means (15) under normal lighting conditions of the scene defining the aforementioned images, the aforementioned first and second compensating values being stored in a non-volatile memory on the aforementioned control means (13); and in the fact that the aforementioned control means are also designed for selecting and modifying the operating coefficients employed by the convoluting means (105).
-
公开(公告)号：EP4435711A1
公开(公告)日：2024-09-25
申请号：EP24160791.0
申请日：2024-03-01
IPC分类号： G06T1/20
CPC分类号： G06T1/20 , G06N3/063 , G06F17/153 , G06N3/045 , G06F17/16
摘要： Methods of implementing a standard deconvolution on a graphics processing unit, the standard deconvolution being representable as a direct convolution between an input tensor to the standard deconvolution and each of a plurality of a sub-filters, each sub-filter of the plurality of sub-filters comprising a subset of weights of a filter of the standard deconvolution. The methods comprising: receiving, at the graphics processing unit, the input tensor in a dense format; identifying, at the graphics processing unit, active positions of the received input tensor; performing, at the graphics processing unit, an indexed unfold operation on the input tensor based on the identified active positions to generate an input matrix comprising elements of the input tensor in each non-zero subwindow of the input tensor; and performing, at the graphics processing unit, a matrix multiplication between a weight matrix and the input matrix to generate an output matrix that comprises elements of an output tensor of the standard deconvolution that are based on the non-zero sub-windows of the input tensor.
-
公开(公告)号：EP4432211A1
公开(公告)日：2024-09-18
申请号：EP24160798.5
申请日：2024-03-01
IPC分类号： G06T1/20
CPC分类号： G06T1/20 , G06N3/063 , G06F17/153 , G06N3/045 , G06F17/16
摘要： Methods of implementing a standard convolution on a graphics processing unit. The methods include: receiving, at the graphics processing unit, an input tensor in a dense format; identifying, at the graphics processing unit, active positions of the input tensor; performing, at the graphics processing unit, an indexed unfold operation on the input tensor based on the identified active positions of the input tensor to generate an input matrix comprising elements of the input tensor in each non-zero window of the input tensor; and performing, at the graphics processing unit, a matrix multiplication between a weight matrix and the input matrix to generate an output matrix that comprises elements of an output tensor of the standard convolution based on the non-zero windows of the input tensor.
-
公开(公告)号：EP4425418A1
公开(公告)日：2024-09-04
申请号：EP24160788.6
申请日：2024-03-01
IPC分类号： G06T1/20
CPC分类号： G06T1/20 , G06N3/063 , G06F17/153 , G06N3/045 , G06F17/16
摘要： Methods of implementing a sparse submanifold convolution using a neural network accelerator. The methods include: receiving, at the neural network accelerator, an input tensor in a sparse format; performing, at the neural network accelerator, for each position of a kernel of the sparse submanifold convolution, a 1x1 convolution between the received input tensor and weights of filters of the sparse submanifold convolution at that kernel position to generate a plurality of partial outputs; and combining appropriate partial outputs of the plurality of partial outputs to generate an output tensor of the sparse submanifold convolution in sparse format.
-
">29.发明授权
公开(公告)号：EP4242855B1
公开(公告)日：2024-06-05
申请号：EP23159428.4
申请日：2023-03-01
CPC分类号： G06F12/0207 , G06F12/0223 , G06F2212/101620130101 , G06F2212/45420130101 , G06F17/153 , G06N3/063 , G06F12/0607 , G06N3/0464 , G06F9/355
-
30.发明公开
公开(公告)号：EP3410307A2
公开(公告)日：2018-12-05
申请号：EP18169744.2
申请日：2018-04-27
CPC分类号： G06F17/153 , G06F17/15 , G06F17/16 , G06F17/175 , G06F17/18
摘要： The present disclosure is applied to a variable group calculation apparatus for calculating an undetermined variable group that simultaneously minimizes a difference value and a data value. The difference value is a difference between an added composite value, which is obtained by adding and combining the undetermined variable group and a dictionary data group, and an observation data group. The data value includes the difference value and a regularization term of the undetermined variable group. The variable group calculation apparatus of the present disclosure includes a convolution unit configured to convert the regularization term to a convolution value for an L1 norm using the undetermined variable group and a mollifier function, and a calculation unit configured to perform the calculation using the regularization term, which is converted to the convolution value by the convolution unit.
-
-
-
-
-
-
-
-
-
搜索结果
99 条记录 (耗时 0.016 秒)
