Binary quantification of an image having multiple levels of grey

Details Binary quantification of an image having multiple levels of grey Binary quantification of an image having multiple levels of grey

1988-05-27

1990-12-11

Moore, David K.

Image analysis

Histogram processing

For setting a threshold

382 52, 382 53, 358466, G06K 938

Patent

active

049776053

DESCRIPTION:

BRIEF SUMMARY
The present invention relates to a process and a device for transforming an initial video-type image, which may include a large number of different grey levels, for example 128 or 256, into a binary image, the darkest image points of the initial image being black on the transformed image, and the lightest image points of the inital image being white on the transformed image.
Such a transformation is necessary in various fields, especially in connection with the collection of information contained in graphics of a more or less dark shade, which graphics are carried by a more or less light support, for example with a view to telecopying, to storage, in electronic form, of documents and plans, to preprocessing for passing into an optical reader of printed characters, or alternatively of manuscript messages, or pre-processing with a view to the recognition of forms in general.
In such a process, for each image point or "pixel" of the initial image it is necessary to determine whether it should correspond to a "white" or "black" pixel of the transformed image.
This may be done by a method which involves the determination of a threshold for the grey levels.
In the simplest concept, the threshold level is fixed in advance, and it is the same for the entire image. This may falsify the message contained in the initial image, in cases where a part of the background, which is considered as light, is darker than the threshold level, or alternatively a part of the graphics is paler than the same threshold level, or vice versa.
It is possible to limit this risk by choosing a different threshold level for each image, this threshold level being determined from the histogram of the grey levels of the initial image. This histogram is usually bimodal, with two distinct maxima, one of which corresponds to the background and the other to the graphics, and the threshold will be fixed at the position of the minimum separating the two maxima of the histogram. However, it happens that this minimum lacks clarity or leads to poor results, particularly when the contrast is low and the background irregular.
It is known to improve this method by breaking down the initial image into a certain number of distinct elements, or windows, and to determine the histogram of the greys and the threshold for each window. This involves a complication of the computation, and reveals anomalies which may interfere with the information, on passing from one window to the other.
Another method does not rely upon the concept of grey intensity threshold, but on the concept of "contour" separating the background from the graphics.
The contour is determined by associating with the initial image a map of the spatial derivatives of the grey intensity. The contour is defined for all the pixels in which this derivative exceeds a minimum value. It is then sufficient to carry out a simple comparison to find whether the interior of a closed contour is to be white or black.
This method is well suited to variable-contrast images, but it necessitates numerous and complex computations.
A need therefore exists for a method which permits the transformation of images with multiple grey levels into a binary image, with a good reliability, even if the contrast is poor, while not necessitating computations as complex as the method indicated herein above.
In order to achieve this object, the invention provides a process for transforming an initial video-type image with multiple grey levels into a binary image, in which the initial image is broken down into a certain number of partial images, a grey level threshold is determined for each partial image and a value corresponding to black or to white of the transformed image is allocated to the central pixel of the partial image, depending upon whether it has a grey level exceeding or below the said threshold. According to the invention, a particular feature of this process is that there is associated with each pixel of the initial image a partial image centered on the said pixel and comprising a plurality of pixels in at least o

REFERENCES:
patent: 4442544 (1984-04-01), Moreland et al.
patent: 4468704 (1984-08-01), Stoffel et al.
patent: 4575768 (1986-03-01), Sakai et al.
patent: 4644410 (1987-02-01), Schlichtig
Hancock, Improved Dynamic Thresholding Algorithm for Image Binarization, 11-78, 2295-2297.
Yasuo Nakagawa et al., Some Experiments on Variable Thresholding, 12-78, 191-204.

Affiliated with

Also associated with

Rating

Binary quantification of an image having multiple levels of grey does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Binary quantification of an image having multiple levels of grey, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Binary quantification of an image having multiple levels of grey will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-394313