JPH07225834A - Picture noise detector - Google Patents

Abstract

PURPOSE:To detect the noise in an unmeasurable area where is not irradiated with pattern light, of a space code picture and that in the other area and to smoothly obtain a reliable value of three-dimensional position information of the surface of a measurement object material. CONSTITUTION:A difference noise detection part 106 which detects the noise in the space code picture obtained from pattern projection pictures, which are obtained by projecting the pattern light having bright and dark area to the object, dependently upon whether the number of difference pictures where difference values don't exceed a threshold D out of plural difference pictures between pattern projection pictures and a reference picture obtained by projecting an arbitrary reference pattern from the same position exceeds a threshold P or not and a dispersed noise detection part 111 which detects the noise in the space code picture in accordance with the variance value and the average value in the vicinity of each point with respect to the space code picture where the noise detected by difference noise detection is corrected are provided.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image noise detecting device.

[0002]

2. Description of the Related Art A plurality of striped pattern lights having different widths are projected on an object, and an image photographed by the camera is binarized,
There is known a technique for obtaining a three-dimensional position information of an object by encoding a space corresponding to each point in an image by a combination of two values in each pattern (3D image measurement by Seiji Iguchi and Kosuke Sato, Shokodo, 1991). The mechanism of this technique will be described with reference to FIGS.

A striped pattern light is projected on an object 405 through a pattern mask 402, which is a projector 401 and has generated a mask pattern according to gray codes 501 to 506. The camera 403 captures an image of an object onto which the striped pattern light is projected. Then, the projected striped pattern is binarized to obtain the value of each bit, whereby the spatial code as indicated by 404 is obtained. The pattern binarization method is a method of binarizing a pattern projection image on which a pattern is projected, an image obtained by averaging an image on which no pattern is projected, and an image on which all patterns of bright regions are projected, from a difference image, A method of binarizing a difference image between a pattern projection image image on which a pattern is projected and an image on which a reverse pattern is projected has been proposed (the above-mentioned document). In this way, the three-dimensional position information of the object surface is measured by the spatial code 404 attached to the surface of the object 405 and the positional relationship between the projector 401 and the camera 403 which is known in advance. Using the spatially coded image obtained on the entire screen of the camera, the three-dimensional position information of the entire range captured on the screen is measured.

According to this three-dimensional information measuring technique, when the spatial code image contains noise, a large error occurs in the three-dimensional position information. Spatial codes are consecutively applied to continuous surfaces, and this continuity is used to make a space, for example, in Japanese Laid-Open Patent Publication No.
It is conceivable to apply a method for detecting noise using an average value and a variance value in the vicinity of each image point, such as the one described in JP-A-3-258166. But this way,
It is not determined whether or not the neighborhood has a reliable value, and when the neighborhood systematically contains noise, a sufficient noise detection effect cannot be obtained. In particular, systematic noise occurs in an unmeasurable area where pattern light does not hit because it is behind an object.

[0005]

SUMMARY OF THE INVENTION An object of the present invention is to provide a device for detecting noise in a spatial code image for obtaining three-dimensional information on the surface of an object.

[0006]

In order to achieve this object, according to the present invention, after the reliability of the neighboring spatial code is evaluated by the variance noise detector in accordance with the magnitude of the variance value near each point of the spatial code image. , It was determined whether or not the point was noise by the value of the difference between the average value of the neighborhood and the value of the point. Further, the difference noise detection unit uses the difference value of the difference image for generating the spatial code image to determine whether or not the point is the unmeasurable region.

[0007]

According to the present invention, in the distributed noise detector,
If the magnitude of the variance value near each point of the spatial code image is larger than a certain value S, it is determined that the spatial code value near the point is systematic noise, and therefore the point is also noise. S
If it is smaller than the above, it is determined that the spatial code value near the point is a reliable value, and if the difference between the average value near the point and the value of the point is large, the point is determined to be noise. By using the median value instead of the average value, the noise at the point can be detected even if some noise is included near the point.

Further, in the difference noise detection unit, when the number of difference images in which the absolute value of the difference value between the pattern projection image and the reference image does not exceed the threshold value D at a certain point exceeds the threshold value P. It is determined that the point is an unmeasurable region where the pattern light does not strike.

[0009]

Embodiments of the present invention will be described in detail below.

(Embodiment 1) A first embodiment of the present invention will be described with reference to FIG.

A pattern projection image obtained by projecting striped pattern light as shown at 501 to 506 in FIG. 5 on an object is stored in a pattern projection image storage unit 102, and a reference image obtained by projecting an arbitrary reference pattern from the same location is stored as a reference image. It is stored in the unit 103. FIG. 6 shows an example of the pattern projection image and the reference image. With respect to the pattern projection image 601, as a reference image,
An image such as 602 that does not project a pattern, an image such as 603 that projects an inverted pattern of 601 can be used. However, in the former case, there is a portion that is a dark pattern regardless of which striped pattern is projected, and the area excluding that portion becomes the measurement target area. The pattern projection image is stored by the number of striped patterns. Further, as described in the section of the related art, the spatial code image obtained based on the pattern projection image is stored in the spatial code image storage unit 101. The difference image calculation unit 104 includes the pattern projection image stored in the pattern projection image storage unit 102 and the reference image storage unit 1.
By subtracting the difference image from the reference image stored in 03,
Obtain the number of pattern projection images.

The difference counting unit 105 examines, for each point of the spatial code image stored in the spatial code image storage unit 101, corresponding points of all the differential images output by the differential image calculation unit 104. If the absolute value of the difference value of the point does not exceed the threshold value D, the point is marked. Then, the number of images in which the point is marked among all the difference images is output as the value of the point.

An example of the processing of the difference counting unit 105 is shown in FIG. Regarding the section indicated by section A (804), the brightness 801 of the pattern projection image, the brightness 802 of the pattern non-projection image serving as the reference image, and the brightness 803 of the reverse pattern projection image also serving as the reference image have the same brightness. Therefore, the surface of the object in that range is not exposed to the pattern light, and the value of the spatial code can be determined to be unreliable.

The section B (805) is a portion where the dark area of the striped pattern light hits, and the point C (806) is a bright area and a dark area where the brightness of the pattern projected image and the inverted pattern projected image are the same. The boundary is shown. All of these are the portions where a reliable spatial code value should be obtained by projecting the pattern, but only the absolute value of the difference value gives a value smaller than the threshold value D, and these points are also the spatial code value. Will be judged as unreliable. The differential noise detection unit 106 functions so as to prevent such a malfunction.

The differential noise detection unit 106, for each point of the spatial code image stored in the spatial code image storage unit 101,
The value output by the difference counting unit 105 is checked, and the number p of difference images in which the absolute value d of the difference value at the point does not exceed the threshold value D
However, if it exceeds the threshold value P, it is determined that the point is in the unmeasurable region where the pattern light does not hit, and the spatial code of the point is noise. When the mask pattern of the striped pattern light is a gray code pattern as shown in FIG. 5, the position of the boundary between the bright region and the dark region is the same in all the difference images output by the difference image calculation unit 104. Therefore, even in the case of the point C (806), the absolute value of the difference value does not decrease in another difference image, and the point C (806) is determined not to be noise. Theoretically, the absolute value d of the difference value should be less than or equal to the threshold value D in all the difference images output by the difference image calculation unit 104 in the area where the striped pattern light is not applied, but the difference value varies. In many cases, the threshold value P is set.
In an experiment conducted by the inventor, d <10 (= D) and p> 4 (= P) for all seven pattern projection images with 256 gradation density.
Detected 95% or more noise.

(Embodiment 2) A second embodiment of the present invention will be described with reference to FIGS.

For each point of the spatial code image stored in the spatial code image storage unit 101, the variance calculation unit 108 calculates the variance value in the vicinity of the point and the average calculation unit 109 calculates the average value in the neighborhood. The inventor uses 8 neighborhoods as shown in FIG. 7 as neighborhoods.

The variance noise detecting unit 111 outputs the variance value output from the variance calculating unit 108 and the average calculating unit 109.
It is determined whether each point is noise by using the average value output by. Table 1 shows the discrimination method.

[0019]

[Table 1]

When the variance value σ of a point is smaller than the threshold value S, the value of the spatial code is stable with little variation in the vicinity of the point, and it can be determined that the value is reliable.
If the absolute value m of the difference between the spatial code value of the point and the average value in the vicinity thereof is smaller than the threshold value M, the point is a stable and reliable value as in the vicinity. However, when m is larger than M, it is determined that the point has a value distant from the neighborhood and is likely to be pinpoint noise. That is, only Mode0 in Table 1 is a reliable value, and it is determined that Mode1 to Mode3 are noises, the noise place is marked, and the result is output. When the object surface is smooth, the space code is given a continuous value, so that the position on the object surface where the pattern light strikes and the three-dimensional position information is stably measured is
Since the spatial code also has continuous values and the variance value is small,
The method described here is effective. As an example, the inventor
= 2.0, M = 4.0 and the experiment is performed.

(Embodiment 3) A third embodiment of the present invention will be described with reference to FIG. As described in the first embodiment, the differential noise detection unit 106 detects noise in the spatial code image stored in the spatial code image storage unit 101. The spatial code image correction unit 107 replaces the noise portion of the spatial code image output by the differential noise detection unit 106 with a value that can distinguish the spatial code value at that point from a normal spatial code value. For example, when the spatial code normally has a value of 0 to 255, the value of the portion determined to be noise is replaced with -1. The spatial code image thus corrected is replaced with the spatial code image stored in the spatial code image storage unit 101 according to the second aspect of the invention. By connecting the invention of claim 1 and the invention of claim 2, the differential noise detector 1 is connected.
The noise that 06 cannot detect due to the difference in the difference value can be detected by the variance noise detection unit 111 by the average value and variance value of the spatial code values in the vicinity thereof.

(Embodiment 4) In the fourth embodiment of the present invention, the average calculator 109 in Embodiments 1 and 3 is replaced with a median calculator 110 which calculates medians of the same neighborhood. As a result, even if some noise is included in the vicinity of a point, the median value can be prevented from fluctuating due to noise, and the noise in the vicinity is not affected by the detection of noise at the point. be able to.

(Fifth Embodiment) As a fifth embodiment of the present invention,
It is characterized in that a pattern obtained by inverting the pattern used for the pattern projection image is used as the reference pattern. Accordingly, the inverted pattern projection image can be used as a common input to the pattern projection image binarization method described in the section of the related art and the differential noise detection unit of the present invention.

[0024]

According to the present invention, 3 by pattern light projection
In the dimensional position information measuring technique, it is possible to detect noise in a non-measurable region where the pattern light does not hit in the spatial code image and noise other than that. By replacing the detected noise location in the spatial code image with a value that can be distinguished from a non-noise location or an interpolated value from the surroundings, a reliable value of the three-dimensional position information of the surface of the measurement target object can be obtained smoothly. It becomes possible.

[Brief description of drawings]

FIG. 1 is a block diagram of the invention of claim 3;

FIG. 2 is a block diagram of the invention of claim 1;

FIG. 3 is a block diagram of the invention of claim 2;

FIG. 4 is a diagram showing a mechanism of a three-dimensional position information measuring technique by pattern light projection.

FIG. 5 is a diagram showing an example of a pattern of patterned light.

FIG. 6 is a diagram showing an example of a pattern projection image and a reference image.

FIG. 7 is a diagram showing eight neighborhoods.

FIG. 8 is a diagram showing an example of processing of a difference counting unit 105.

[Description of Reference Signs] 101 spatial code image storage unit 102 pattern projection image storage unit 103 reference image storage unit 104 difference image calculation unit 105 difference counting unit 106 differential noise detection unit 107 spatial code image correction unit 108 variance calculation unit 109 average calculation unit 110 Median Value Calculation Unit 111 Variance Noise Detection Unit 401 Projector 402 Pattern Mask 403 Camera 404 Spatial Code 405 Objects 501 to 506 Mask Pattern Example 601 Pattern Projected Image Example 602 Reference Image Example (No Pattern Projected Image) 603 Reference Image Example (reverse pattern projection image) 701 to 708 8 neighborhoods of a certain point 709 The point 801 Graph of pattern projection image 802 Graph of no pattern projection image 803 Graph of reverse pattern projection image 804 Section A 805 Section B 806 Point C

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Office reference number FI Technical display location H04N 1/409 7459-5L G06F 15/70 330 Z H04N 1/40 101 C

Claims (5)

[Claims] 1. A pattern projection image storage unit for storing a plurality of pattern projection images obtained by projecting striped pattern light composed of a bright region and a dark region on an object, and from the same location where the pattern light is projected. A reference image storage unit for storing one or a plurality of reference images projected with an arbitrary reference pattern;
A difference image calculation unit that obtains a difference image by subtracting the pattern projection image and the reference image, and a spatial code image storage that stores a spatial code image that encodes a space in an image obtained from the plurality of pattern projection images. A difference counting unit that counts the number of copies of the plurality of difference images in which the absolute value of the difference value corresponding to each point does not exceed the threshold value D.
A difference noise detection unit for detecting whether or not the spatial code of the point stored in the spatial code image storage unit is noise depending on whether or not the value corresponding to each point of the difference counting unit exceeds a threshold value P. An image noise detection device having. 2. A spatial code image storage unit for storing a spatial code image in which a space in an image is coded, and the spatial code image,
A variance calculation unit that calculates a variance value in the vicinity of each point, an average calculation unit that calculates an average value in the vicinity of each point of the spatial code image,
From the value of the difference between each point and the average value in the vicinity of each point, and from the variance value in the vicinity of each point, a noise detection unit that detects whether or not each point of the spatial code image is noise is provided. A characteristic image noise detection device. 3. A spatial code image correction unit for replacing a value of a spatial code with a value that can be distinguished from other non-noise values for a location where noise is detected by the differential noise detection unit according to claim 1. The image noise detection apparatus according to claim 1, wherein the code image correction unit is replaced with the spatial code image storage unit according to claim 2. 4. The image noise detection apparatus according to claim 2, wherein the average calculation unit is replaced with a median value calculation unit that calculates a median value near each point of the spatial code image. 5. A reference pattern projected onto an object comprises bright and dark regions obtained by inverting the striped pattern light projected onto the object in order to obtain a pattern projection image. Alternatively, the image noise detection device according to item 4.