News Column

Patent Issued for Data Processing Device, Image Matching Method, Program, and Image Matching System

June 24, 2014



By a News Reporter-Staff News Editor at Information Technology Newsweekly -- A patent by the inventor Hamada, Yasushi (Tokyo, JP), filed on March 23, 2010, was published online on June 10, 2014, according to news reporting originating from Alexandria, Virginia, by VerticalNews correspondents.

Patent number 8749658 is assigned to NEC Corporation (Tokyo, JP).

The following quote was obtained by the news editors from the background information supplied by the inventors: "Generally, image matching is used to identify a person and letters based on image information. In image matching, a target image in which a target object is captured and a reference image in which the object has been captured and stored are compared for similarity so as to identify a person or letters contained in the captured target image. Images are compared for similarity generally according to SAD (Sum of Absolute Difference), SSD (Sum of Squared Difference), ZNCC (Zero-mean Normalized Cross-Correlation), or the like.

"These techniques compare luminance values of a target image and a reference image. Thus, when images of an object are captured, if illumination conditions of the target image and the reference image differ, the luminance values of these images will vary and thereby the accuracy of matching tends to deteriorate.

"Non-patent Literature 1 proposes a technique called increment sign correlation that performs robust image matching under illumination variation. The increment sign correlation method extracts only the sign of increment (relationship of magnitude) of luminance values of adjacent pixels in the horizontal direction for each of a target image and a reference image. Let f(x, y) as the luminance value of coordinates (x, y), an increment sign value h(x, y) is defined as Formula (1).

".times..times..times..times..function..times..times..function..gtoreq..fu- nction. ##EQU00001##

"In the increment sign correlation, the number of pixels of which increment sign value of a target image matches that of a reference image is evaluated to match the images.

"When illumination varies, the luminance value of each pixel and contrast (luminance difference) vary. Since the increment sign value represents the sign of the increment (relationship of magnitude) of luminance values of adjacent pixels, the increment sign value is hardly influenced by variation in the luminance value and contrast due to illumination variation.

"Patent Literature 1 also proposes a technique that matches images for a target object using a plurality of reference images so as to improve the accuracy of matching. In Patent Literature 1, the increment sign values are obtained for each of a plurality of reference images and the probability of which the increment sign value becomes 1 is computed for each pixel. An image is generated based on computed probability values as pixel values and then the generated image and the target image are matched.

"Patent Literature 2 presents a method that detects eyes from a facial image."

In addition to the background information obtained for this patent, VerticalNews journalists also obtained the inventor's summary information for this patent: "Problem to be Solved by the Invention

"In the above-described Patent Literature 1, image matching tends to be easily influenced by shading due to the shape of a target object. FIG. 1 is a schematic diagram showing lettered spherical objects illuminated in a different way and their increment sign values. The upper left illustration shown in FIG. 1 represents an image of which an object is illuminated in front of the right, whereas the lower left illustration represents an image of which an object is illuminated in front of the left.

"When the image of which the object is illuminated in front of the right is converted into increment sign values defined as Formula (1), an image shown in the upper right illustration is obtained; when the image of which the object is illuminated in front of the left is converted into increment sign values defined as Formula (1), an image as shown in the lower right illustration is obtained.

"When these images are converted into increment sign values defined as Formula (1), if the luminance values increase from the left to the right of the image (the image becomes brighter from the left to the right), the increment sign values become 1; when the luminance values decrease from the left to the right (the image becomes darker from the left to the right), the increment sign values become 0. By showing the regions having an increment sign value of 1 as black and the regions having an increment sign value of 0 as white, when the image of which the object is illuminated in front of the right is converted into increment sign values, the image shown in the upper right illustration is obtained, when the image of which the object is illuminated in front of the left is converted into increment sign values, the image shown in the lower right illustration is obtained.

"In other words, since increment sign values represent relative luminance in a object surface, as shown in FIG. 1, if the object is three-dimensional shape, increment sign values represent different characteristics depending on illumination direction. In a region that has a larger contrast than influenced by illumination variation, for example a letter pattern region, increment sign values represent the same characteristics. However, since a flat region with no letter patterns is strongly influenced by illumination variation, even if the shape of the object is the same, the increment sign values represent different characteristics. Thus, images having many flat regions and few characteristic patterns such as letter patterns are likely to be incorrectly matched.

"For example, in the case of facial image matching, since it has larger non-characteristic regions such as forehead and cheeks than regions having characteristic patterns such as eyes, nose, and mouth, the matching is strongly influenced by illumination variation.

"In addition, in the above-described Patent Literature 1, to improve the accuracy of matching, many reference images need to be collected for each target object. However, it is difficult to equally collect many reference images. To do that, it takes a long time and knowhow to initially set up the image matching system. In particular, in the case of facial image matching, many reference images need to be collected for each of the parts such as eyes, nose, and mouth. Thus, difficulty will rise.

"A purpose of the present invention is to provide a technique that allows robust image matching to be easily performed under illumination variation.

"Means that Solve the Problem

"To accomplish the above-described object, a data processing device according to the present invention comprises:

"difference value computing means that computes a difference value between a pixel value of a target pixel that is each pixel contained in an image and a pixel value of a pixel that is present at a predetermined neighboring relative position of the target pixel;

"representative value computing means that extracts a pixel group containing pixels that are similarly influenced by shading due to light from the image with respect to each pixel of the image and that computes a representative value of difference values of the pixel group according to a statistical technique;

"feature value computing means that computes the feature value with respect to each pixel contained in the image based on comparison between the difference value with respect to each pixel and the representative value of difference values of the pixel group; and

"similarity determining means that determines a similarity between the image and a predetermined image based on the feature value with respect to each pixel extracted by the feature value computing.

"In addition, to accomplish the above-described object, an image matching method according to the present invention comprises:

"computing a difference value between a pixel value of a target pixel that is each pixel contained in an image and a pixel value of a pixel that is present at a predetermined neighboring relative position of the target pixel;

"extracting a pixel group containing pixels that are similarly influenced by shading due to light from the image with respect to each pixel of the image and computing a representative value of difference values of the pixel group according to a statistical technique;

"computing the feature value with respect to each pixel contained in the image based on comparison between the difference value with respect to each pixel and the representative value of difference values of the pixel group; and

"determining a similarity between the image and a predetermined image based on the feature value with respect to each pixel.

"In addition, to accomplish the above-described object, a program according to the present invention is a program of a data processing device that matches images, the program causing a computer to execute processes comprising:

"computing a difference value between a pixel value of a target pixel that is each pixel contained in an image and a pixel value of a pixel that is present at a predetermined neighboring relative position of the target pixel;

"extracting a pixel group containing pixels that are similarly influenced by shading due to light from the image with respect to each pixel of the image and computing a representative value of difference values of the pixel group according to a statistical technique;

"computing the feature value with respect to each pixel contained in the image based on comparison between the difference value with respect to each pixel and the representative value of difference values of the pixel group; and

"determining a similarity between the image and a predetermined image based on the feature value with respect to each pixel.

"In addition, to accomplish the above-described object, an image processing system according to the present invention comprises:

"an image capturing device that captures a target object and outputs a captured image;

"a storage device that stores a reference image of the target object that has been captured; and

"a data processing device that computes a difference value between a pixel value of a target pixel that is each pixel contained in an image and a pixel value of a pixel that is present at a predetermined neighboring relative position of the target pixel for each of the captured image and the reference image, extracts a pixel group containing pixels that are similarly influenced by shading due to light from the image with respect to each pixel of the image and computing a representative value of difference values of the pixel group according to a statistical technique for each of the captured image and the reference image, computes the feature value with respect to each pixel contained in the image based on comparison between the difference value with respect to each pixel and the representative value of difference values of the pixel group for each of the captured image and the reference image, and determines a similarity between the captured image and the reference image based on the feature value with respect to each pixel.

"Effect of the Invention

"According to the present invention, robust image matching can be easily performed under illumination variation."

URL and more information on this patent, see: Hamada, Yasushi. Data Processing Device, Image Matching Method, Program, and Image Matching System. U.S. Patent Number 8749658, filed March 23, 2010, and published online on June 10, 2014. Patent URL: http://patft.uspto.gov/netacgi/nph-Parser?Sect1=PTO1&Sect2=HITOFF&d=PALL&p=1&u=%2Fnetahtml%2FPTO%2Fsrchnum.htm&r=1&f=G&l=50&s1=8749658.PN.&OS=PN/8749658RS=PN/8749658

Keywords for this news article include: NEC Corporation, Information Technology, Information and Data Processing.

Our reports deliver fact-based news of research and discoveries from around the world. Copyright 2014, NewsRx LLC


For more stories covering the world of technology, please see HispanicBusiness' Tech Channel



Source: Information Technology Newsweekly


Story Tools






HispanicBusiness.com Facebook Linkedin Twitter RSS Feed Email Alerts & Newsletters