Image processing apparatus using neural network
First Claim
1. An image processing apparatus for estimating multivalued image data from binary image data converted by a binarizing method which exerts a large influence on surrounding multivalued pixels in specific directions with respect to a multivalued pixel undergoing binarization by the method, comprising:
- input means for inputting binary image data comprising a plurality of pixels which include a pixel of interest that is to be subjected to multivalued conversion, the plurality of pixels being arranged so that a number of binary pixels in a first specific direction with respect to the position of the pixel of interest is larger than that of pixels in a direction other than the first specific direction, the pixels in the first specific direction being ones having a large influence upon the pixel of interest; and
multivalued converting means for executing processing, using a neural network, to restore the input binary image data to multivalued image data for the pixel of interest, wherein said neural network comprises;
an input layer which includes a plurality of neurons for inputting the binary image data from the input means in individual pixel units;
an intermediate layer which includes a plurality of neurons, connected to the input layer, for receiving outputs from the neurons of the input layer; and
an output layer having one neuron, connected to the intermediate layer, for receiving outputs from the neurons of the intermediate layer and outputting one multivalued image data as image data of the pixel of interest.
1 Assignment
0 Petitions
Accused Products
Abstract
Disclosed is an image processing apparatus having an input device for inputting binary image data comprising a plurality of pixels which include a pixel of interest that is to be subjected to multivalued conversion, the plurality of pixels being contained in an area that is asymmetrical with respect to the position of the pixel of interest, and an multivalued converting device for executing processing, by a neural network, to restore the input binary image data to multivalued image data for the pixel of interest, whereby multivalued image data is estimated from binarized image data. It is possible to reduce the number of pixels referred to in arithmetic operations performed in the neural network.
-
Citations
22 Claims
-
1. An image processing apparatus for estimating multivalued image data from binary image data converted by a binarizing method which exerts a large influence on surrounding multivalued pixels in specific directions with respect to a multivalued pixel undergoing binarization by the method, comprising:
-
input means for inputting binary image data comprising a plurality of pixels which include a pixel of interest that is to be subjected to multivalued conversion, the plurality of pixels being arranged so that a number of binary pixels in a first specific direction with respect to the position of the pixel of interest is larger than that of pixels in a direction other than the first specific direction, the pixels in the first specific direction being ones having a large influence upon the pixel of interest; and multivalued converting means for executing processing, using a neural network, to restore the input binary image data to multivalued image data for the pixel of interest, wherein said neural network comprises; an input layer which includes a plurality of neurons for inputting the binary image data from the input means in individual pixel units; an intermediate layer which includes a plurality of neurons, connected to the input layer, for receiving outputs from the neurons of the input layer; and an output layer having one neuron, connected to the intermediate layer, for receiving outputs from the neurons of the intermediate layer and outputting one multivalued image data as image data of the pixel of interest. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10)
-
-
11. An image processing apparatus for estimating multivalued image data from binarized image data, comprising:
-
input means for inputting binary image data comprising a plurality of pixels which include a pixel of interest that is to be subjected to multivalued conversion, the plurality of pixels being contained in an area that is asymmetrical with respect to the position of the pixel of interest; and multivalued converting means for executing processing, using a neural network, to restore the input binary image data to multivalued image data for the pixel of interest, wherein said multivalued converting means further comprises; discriminating means for discriminating the inputted binary image data by a first neural network to determine whether the image data is indicative of a fine-line and a character portion or of an image portion; first restoring means for applying multivalued restoration processing, by a second neural network, to the fine-line and the character portions contained in the input binary image data; second restoring means for applying multivalued restoration processing, by a third neural network to the image portion contained in the input binary image data; and control means or controlling the multivalued restoration processing to be executed by a proper one of said first or said second restoring means, based upon the results of discrimination by said discriminating means. - View Dependent Claims (12, 14, 15, 16, 17, 18, 19, 20, 21, 22)
-
-
13. An image processing method for a neural network of estimating multivalued image data from binary image data converted by a binarizing method which exerts a large influence on surrounding multivalued pixels in specific directions with respect to a multivalued pixel under binarization by the method, said method comprising the steps of:
-
inputting binary image data to an input layer of the neural network in individual pixel units, the binary image data comprising a plurality of pixels which include a pixel of interest that is to be subjected to multivalued conversion, the plurality of pixels being arranged so that a number of binary pixels in a first specific direction with respect to the position of the pixel of interest is larger than that of pixels in a direction other than the first specific direction, the pixels in the first specific direction being ones having a large influence upon the pixel of interest; processing data output from the input layer in an intermediate layer of the neural network which includes a plurality of neurons; processing data output from the intermediate layer in an output layer in the neural network which includes one neuron; and outputting one multivalued data from the output layer as a multivalued image data of the pixel of interest.
-
Specification