Image processing apparatus for recognition of linear structures

摘要

An image process apparatus is operative to obtain an image with high visual recognition property. A linear structural object incorporated into the original image is distinguished by two methods. A first method produces an evaluation image (P3) that evaluates whether each pixel is a linear structural object in the original image. A second method produces the difference image (P6) incorporating a linear structural object by obtaining the difference between the linear structural object incorporated into the original image and the portion other than the linear structural object. Since the linear structural object in the original image is extracted from an original image (P0) holding the contrasting density in the original image based on the two images related to the linear structural object produced by such different methods, and the apparatus provides an image having high visual recognition property.

主权项

1. An image processing apparatus, that processes an original image containing a plurality of pixels obtained by sight-through of a subject, comprising:
an evaluation image producing circuit operative to produce an evaluation image that evaluates whether or not each said pixel in said original image is a linear structural object in said original image based only on said original image, whereby said evaluation image further contains information about a position of said linear structural object in said original image; a direction image producing circuit operative to produce a direction image that indicates an extending direction to which a line extends from each said pixel in said original image; a difference image producing circuit operative to produce a difference image from said original image in which a pixel value in a proximity of said linear structural object of the original image is subtracted from a pixel value of a portion of said linear structural object based upon said direction image; an extraction image producing circuit operative to produce an extraction image in which the linear structural object in the original image is extracted from the original image based on said evaluation image and said difference image; and whereby a location of said linear structural object in said original image is obtained accurately and at high-speed in real time from said original image even under a noisy image environment.