Title:
【発明の名称】ニューラルネットによる画像の2値化処理
Document Type and Number:
Japanese Patent JP3386822
Kind Code:
B2
Abstract:
PURPOSE:To binarize the gradation image by a neural network. CONSTITUTION:By allowing an average of a gradation value of a binary picture in the vicinity of a certain pixel to approach an original gradation value of its pixel, a binary image of a high quality is obtained. Neurons which are asynchronous but are separated enough can be operated simultaneously, therefore, can be operated at a high speed. Each neuron inputs a state (1 or 0) of peripheral (d) points, compares its sum and a threshold and determines a state. By operating asynchronously all the neurons, the whole network is stabilized securely. Even if plural neurons separated enough are operated simultaneously (determinately), the same result as they are operated one by one simultaneously is obtained. Accordingly, by operating simultaneously only one neuron in the vicinity and operating simultaneously plural neurons as a whole, the network can be stabilized at a high speed.
Inventors:
Taizo Hattori
Munemitsu Ikegami
Tanaka
Munemitsu Ikegami
Tanaka
Application Number:
JP351491A
Publication Date:
March 17, 2003
Filing Date:
January 16, 1991
Export Citation:
Assignee:
Tanaka
International Classes:
G06F15/18; G06N3/00; G06N99/00; G06T1/00; H04N1/40; H04N1/403; (IPC1-7): G06T1/00; G06F15/18; G06N3/00; H04N1/403
Domestic Patent References:
| JP272491A | ||||
| JP57104369A | ||||
| JP2179166A |
Attorney, Agent or Firm:
Noriyasu Sakamoto