Appeal No. 1999-1587
Application 08/601,751
The disclosed invention pertains to a neurofilter and
a method of training the neurofilter to operate on image data
so as to discriminate between text and picture regions of an
image which is expressed by image data.
Representative claims 6 and 9 are reproduced as
follows:
6. A neurofilter comprising:
a neural network having weighting coefficients which
may be adjusted during a training procedure, said neural
network having been subjected to a training procedure for
setting said weighting coefficients such as to provide a
specific type of filtering of image data or serial data which
are sequentially supplied to said neural network,
wherein said neurofilter functions as a separation
filter for operating on image data expressing an original
image, said neurofilter responding to data of respectively
different predetermined types of image region within said
original image by producing an output signal at
correspondingly different levels,
said neurofilter further comprising
smoothing filter means for executing smoothing
processing of said output signal.
9. A filter apparatus comprising a parallel combination
of a conventional filter and a neurofilter, having respective
inputs thereof coupled in common, and means for combining
output data produced from said neurofilter with output data
produced from said conventional filter to thereby compensate
for errors in said output data from the conventional filter,
said neurofilter comprising a neural network having weighting
coefficients which may be adjusted during a training
-2-
Page: Previous 1 2 3 4 5 6 7 8 9 10 11 Next
Last modified: November 3, 2007