Edge extracting method and apparatus using diffusion neural network
First Claim
1. An apparatus for extracting edges of an object from an image signal, comprising:
- a distribution neural network including a plurality of neurons, each neuron inducing an excitatory response based on a first predetermined connection weight value according to an intensity of said image signal, and inducing an inhibitory response based on a second predetermined weight value according to said intensity of said image signal; and
a diffusion neural network including a plurality of neurons for forming a Gaussian distribution representing a regularity of the excitatory and inhibitory responses, each neuron of said diffusion neural network superposing said excitatory response induced by a corresponding neuron of said distribution neural network and said inhibitory responses induced by neurons of said distribution neural network, which are adjacent to said corresponding neuron of said distribution neural network, and diffusing a superposed signal generated by said superposing to form said Gaussian distribution, and the diffusion neural network convolving the Gaussian distribution and the image signal to detect edges of an object in an image represented by said image signal.
1 Assignment
0 Petitions
Accused Products
Abstract
Edge extracting method and apparatus using a neural network performing a function of diffusing an excitation. The method and apparatus continuously detect a variety of intensity changes of an image via a function having a variety of frequency characteristics. An edge of a fixed object is detected from images continuously input, and an edge of a moving object is selectively detected from the images. The edge extracting apparatus includes a first neural network which receives an image signal. The first neural network derives a Gaussian function representing the regularity of an excitatory response and an inhibitory response to a spot excitation of the image signal. The apparatus also includes a second neural network which detects edges of an image represented by the image signal by convolving the Gaussian function and the image signal.
18 Citations
17 Claims
-
1. An apparatus for extracting edges of an object from an image signal, comprising:
-
a distribution neural network including a plurality of neurons, each neuron inducing an excitatory response based on a first predetermined connection weight value according to an intensity of said image signal, and inducing an inhibitory response based on a second predetermined weight value according to said intensity of said image signal; and a diffusion neural network including a plurality of neurons for forming a Gaussian distribution representing a regularity of the excitatory and inhibitory responses, each neuron of said diffusion neural network superposing said excitatory response induced by a corresponding neuron of said distribution neural network and said inhibitory responses induced by neurons of said distribution neural network, which are adjacent to said corresponding neuron of said distribution neural network, and diffusing a superposed signal generated by said superposing to form said Gaussian distribution, and the diffusion neural network convolving the Gaussian distribution and the image signal to detect edges of an object in an image represented by said image signal. - View Dependent Claims (2, 3, 4, 5, 6)
-
-
7. An apparatus for extracting edges of an object from an image signal, comprising:
-
a distribution neural network including a plurality of neurons, each neuron inducing an excitatory response based on a first predetermined connection weight value according to an intensity of said image signal, and inducing an inhibitory response based on a second predetermined weight value according to said intensity of said image signal; and a diffusion neural network including a plurality of neurons for forming a Gaussian distribution representing a regularity of the excitatory and inhibitory responses, each neuron of said diffusion neural network superposing said excitatory response induced by a corresponding neuron of said distribution neural network and said inhibitory responses induced by neurons of said distribution neural network, which are adjacent to said corresponding neuron of said distribution neural network, and diffusing a superposed signal generated by said superposing to form said Gaussian distribution, and the diffusion neural network convolving the Gaussian distribution and the image signal; a delay means for delaying an output signal from said diffusion neural network; and a differential means for obtaining a difference between said output signal from said diffusion neural network and a signal output from said delay means to detect edges of a moving object in an image represented by said image signal. - View Dependent Claims (8)
-
-
9. A method for extracting an edge from an image signal, comprising the steps of:
-
(a) receiving, with a first neural network including a plurality of neurons, said image signal; (b) each neuron of said first neural network inducing an excitatory response based on a first predetermined connection weight value according to an intensity of said image signal; (c) each neuron of said first neural network inducing an inhibitory response based on a second predetermined connection weight value according to said intensity of said image signal; (d) deriving, with a second neural network including a plurality of neurons, a Gaussian distribution representing a regularity of the excitatory and inhibitory responses, wherein each neuron of said second neural network superposes said excitatory response induced by a corresponding neuron of said first neural network and said inhibitory responses induced by neurons of said first neural network, which are adjacent to said corresponding neuron of said first neural network, and diffusing a superposed signal generated by said superposing; (e) convolving said Gaussian distribution and said image signal; and (f) detecting edges of a fixed object from a signal output at said step (e). - View Dependent Claims (10, 11, 12, 13, 14, 15)
-
-
16. A method for extracting an edge from an image signal, comprising the steps of:
-
(a) receiving, with a first neural network including a plurality of neurons, said image signal; (b) each neuron of said first neural network inducing an excitatory response based on a first predetermined connection weight value according to an intensity of said image signal; (c) each neuron of said first neural network inducing an inhibitory response based on a second predetermined connection weight value according to said intensity of said image signal; (d) deriving, with a second neural network including a plurality of neurons, a Gaussian distribution representing a regularity of the excitatory and inhibitory responses, wherein each neuron of said second neural network superposes said excitatory response induced by a corresponding neuron of said first neural network and said inhibitory responses induced by neurons of said first neural network, which are adjacent to said corresponding neuron of said first neural network, and diffusing a superposed signal generated by said superposing; (e) convolving said Gaussian distribution and said image signal; (f) delaying an output signal from said step (e); and (g) obtaining a difference between said output signal at said step (e) and a signal output at said step (f) to detect edges of a moving object in an image represented by said image signal. - View Dependent Claims (17)
-
Specification