JP2010093736A - Noise canceling method, and coating checking method and apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a noise cancellation processing method reduced in influence on an image portion required for a user. <P>SOLUTION: A noise canceling method has: a step of capturing image data, containing a density for each pixel, obtained by an imaging apparatus; a dissimilar density average calculation step of calculating a moving average of dissimilar densities from the density for each pixel; and a differential data calculation step of subtracting the moving average of the dissimilar densities from the density for each pixel to determine a difference between both moving averages. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a technique for removing noise from an image signal, and further relates to a coating inspection technique using a noise removal technique.

  The image signal obtained from the image sensor inevitably contains noise. Conventionally, various methods for removing noise from an image signal are known.

  Patent Document 1 describes that an image signal is separated into a luminance signal and a color difference signal, and noise removal of the luminance signal and noise removal of the color difference signal are performed separately. Patent Document 2 describes a noise removal method focusing on fixed pattern noise among fixed pattern noise and random noise. Patent Document 3 describes an image processing apparatus that performs digital dodging processing. The digital dodging process is an application of the dodging process in analog photographic technology, that is, a process of covering a part of an image to obtain a print with appropriate brightness at the time of printing.

JP 2006-309524 A JP 2004-015712 A JP 2006-330938 A

  In general, in the noise removal process, the image signal is smoothed. Therefore, noise removal processing removes noise, but normal image signals other than noise are also smoothed. For example, some digital images are composed of an image portion that the user wants to leave and an image portion that the user does not want to leave depending on the application. When noise removal processing is performed on such an image, the image portion that the user wants to leave is also affected.

  An object of the present invention is to provide a noise removal processing method that has little influence on an image portion necessary for a user.

  According to the present invention, the noise removal method includes a step of obtaining image data including the density for each pixel obtained by the imaging device, and a non-similarity density calculating the moving average of the dissimilar density from the density for each pixel by the arithmetic processing unit. A similar density average calculating step; and a difference data calculating step of subtracting the moving average of the dissimilar density from the density of each pixel by the arithmetic processing unit to obtain a difference between the two.

The dissimilar density calculation step includes the following steps.
First, pixels are sequentially extracted from the pixels constituting the image data at a predetermined pitch, and are set as reference pixels bi (i = 1 to k).

Next, for each of the reference pixels bi (i = 1 to k), a predetermined area including the reference pixel is set, and is set as a calculation area Bi.
Next, for each of the calculation areas Bi, a predetermined number of pixels having a large difference from the density qi of the reference pixel bi are extracted from the pixels included in the calculation area.

Next, the average value cmi of the density of the extracted pixels is calculated for each of the calculation areas Bi.
Finally, each density qi of the reference pixel is replaced by each of the average values cmi of the dissimilar densities.

  According to the present invention, it is possible to remove noise in an unnecessary image portion without affecting the image portion necessary for the user.

  With reference to FIG. 1, the outline of the inspection method of the coating of the vehicle body of a motor vehicle is demonstrated. The coating inspection apparatus includes an illumination device 11 that irradiates an automobile body 10 with illumination light, an optical system (lens unit) 12 that enlarges or reduces an image of the vehicle body, a camera 13 that captures a digital image of the vehicle body, and an image of the vehicle body. An image processing device 14 for detecting coating defects. The camera 13 may be a CCD camera. The image processing device 14 may be a normal computer including a storage device, an arithmetic processing device, and a memory. Such a computer includes an input device, a display device, and a printer. The coating inspection apparatus of the present example is configured to divide the surface of the body of an automobile into small regions and automatically perform coating inspection for each region. The coating inspection apparatus moves around the vehicle body by a robot (not shown).

  With reference to FIG. 2, the procedure of a coating defect detection process in the image processing apparatus will be described. First, in step S101, the arithmetic processing device provided in the image processing device 14 takes in image data from the camera and stores it in a memory or a storage device. In step S102, the arithmetic processing unit performs differentiation processing on the image data. Thereby, the density change of the image data is obtained. The region where the density change is large may be a defect, but may be other than a defect. The density change is large at the edges of doors, door knobs, window glass, headlamps, and the like. In step S103, the arithmetic processing unit performs noise removal processing on the image data. Noise removal processing is smoothing of image data. The noise removal process will be described in detail below with reference to FIG. In step S104, the arithmetic processing device binarizes the image data. As a result, the image is divided into two areas, a high density area and a low density area. For example, the image is divided into black and white. A defect candidate can be detected from this binary image. For example, the white area is determined to be a defect candidate. In step S105, the arithmetic processing unit performs a mask process. That is, the mask area is removed from the binary image. Door knobs, window glass, headlamps, etc. are mask areas because they are not painted. Finally, in step S106, the arithmetic processing unit detects a defect. Of the defect candidates detected in step S104, those other than the mask area are determined to be paint defects.

  FIG. 6 shows an example of an original image 601 captured from a camera and an image 602 that has been subjected to binarization processing after noise removal. The original image 601 includes noise 6012 and an image region that the user wants to leave, that is, a coating defect 6013. Noise 6012 is represented by small gray circles and paint defects 6013 are represented by white circles. According to the present invention, in the image 602 after the noise removal process, the noise 6012 is removed from the original image 601 and only the coating defect 6023 remains. That is, the noise is removed, but the image that the user wants to leave is clearly left without deterioration.

  The noise removal method according to the present invention will be described with reference to FIG. In step S201, the arithmetic processing device provided in the image processing device 14 takes in the image data and stores it in the memory or the storage device. This image data may be image data captured from the imaging device. With this image data, for example, an original image 601 shown in FIG. 6 is obtained. In step S202, the arithmetic processing unit calculates a moving average of the density of the pixel data. A method for calculating the moving average of the density will be described later in detail with reference to FIG. By obtaining the moving average of the density, the density distribution of the original image 601 is smoothed. In step S203, the arithmetic processing unit calculates a moving average of dissimilar densities. The method for calculating the dissimilar density average will be described in detail later with reference to FIG. By obtaining the dissimilar density average, the density of the original image 601 is inverted. That is, the density is low in the high density area, and the density is high in the low density area. In particular, in the region where the density change is large, the inversion of the density becomes remarkable. Finally, in step S204, the arithmetic processing unit subtracts the moving average of the dissimilar density from the image data of the original image to obtain the difference between the two. In the difference data obtained in this way, noise in the original image 601 is smoothed or removed, and only the coating defect remains without being smoothed. In the noise removal method according to the present invention, the process of calculating the moving average of the density of the pixel data in step S202 may be omitted. In this case, noise is removed only by step S201, step S203, and step S204.

  Graphs 701 to 704 in FIG. 7 show the density distribution of the image. In these graphs, the horizontal axis represents pixel positions along one scanning line, and the vertical axis represents density values. A curve 7011 of the graph 701 in FIG. 7 indicates the density distribution of the original image 601 captured in step S201. Peak 7012 represents noise and peak 7013 represents a paint defect. A curve 7021 in the graph 702 in FIG. 7 shows the moving average distribution of the density calculated in step S202. By obtaining the moving average of the density, the density distribution of the original image is smoothed. A curve 7031 in the graph 703 in FIG. 7 shows the moving average distribution of dissimilar densities calculated in step S203. By calculating the dissimilar density average, the peak unevenness in the curve 7011 representing the density distribution of the original image 601 is inverted. A curve 7041 of the graph 704 in FIG. 7 indicates the distribution of differences calculated in step S204. In the difference data, the peak 7012 indicating noise in the curve 7011 representing the density distribution of the original image 601 is smoothed or removed, and the coating defect 7013 remains without being smoothed. Therefore, when binarization is performed using the data of the curve 7041 of the graph 704 in FIG. 7 and then imaged, the noise 6012 is removed as shown in the image 602 in FIG. Remains.

  With reference to FIG. 4, the method for calculating the moving average of the density in step S202 of FIG. 3 will be described. In step S301, the arithmetic processing device provided in the image processing device 14 takes in image data that is a target of calculation of a moving average of density and stores it in a memory or a storage device. In this example, the image data of the original image 601 captured from the camera, that is, the original image data. From the pixels constituting the image data, pixels are sequentially extracted at a predetermined pitch. This is defined as a reference pixel ai (i = 1 to k). The density of these reference pixels is pi (i = 1 to k). If the pitch is one pixel, all pixels are taken out in order. The order of taking out the pixels may be the scanning order of the pixels constituting the pixel data.

  In step S302, the arithmetic processing unit sets a predetermined region including the reference pixel for each reference pixel thus extracted. This is defined as a calculation area Ai (i = 1 to k). The number of pixels included in the calculation area Ai is preferably the same in all calculation areas. However, when the reference pixel is present at the edge of the image, the number of pixels included in the calculation region may be smaller. In step S303, the arithmetic processing unit calculates an average value of the densities of the pixels included in each calculation region. This is the average density pmi (i = 1 to k). In step S304, the arithmetic processing unit replaces the density pi of the reference pixel with the density average value pmi. Such calculation of the average value of density and its replacement are sequentially performed for all the reference pixels. Thereby, a moving average of the concentration is obtained.

  With reference to FIG. 5, the method of calculating the dissimilar density average in step S203 of FIG. 3 will be described. In step S401, the arithmetic processing device provided in the image processing device 14 takes in the image data that is the target of calculation of the moving average of the dissimilar density, and stores it in the memory or the storage device. In this example, this image data is a moving average of density calculated in step S202. The arithmetic processing device extracts pixels one by one from the pixels constituting the image data in order at a predetermined pitch. This is set as a reference pixel bi (i = 1 to k). The density of these reference pixels is qi (i = 1 to k). If the pitch is one pixel, all pixels are taken out in order. The order of taking out the pixels may be the scanning order of the pixels constituting the pixel data. In step S402, the arithmetic processing unit sets a predetermined region including the reference pixel for each reference pixel bi (i = 1 to k). This is a calculation area Bi (i = 1 to k). However, the calculation area Bi set here is preferably larger than the calculation area Ai set in step S302. The number of pixels included in the calculation area Bi is preferably the same in all calculation areas. However, when the reference pixel is present at the edge of the image, the number of pixels included in the calculation region may be as small as possible.

  In step S403, the arithmetic processing unit extracts a predetermined number of pixels having a large difference from the density qi of the reference pixel bi among the pixels included in each calculation region Bi in descending order of the density difference. That is, a pixel having a density that is not similar to the density qi of the reference pixel bi is taken. For example, it is assumed that n pixels cij (j = 1 to n, where n is smaller than the number of pixels included in the region Bi) are extracted. In step S404, the arithmetic processing unit calculates an average value of the densities of n pixels cij (j = 1 to n). Let the average value of this density | concentration be cmi (i = 1-k). This is the average value of dissimilar densities. In step S405, the arithmetic processing unit replaces the density qi of the reference pixel bi with the average value cmi of the dissimilar density. The calculation of the average value of such dissimilar densities and the replacement thereof are sequentially performed for all the reference pixels. Thereby, a moving average of dissimilar concentrations is obtained.

  The average value of dissimilar densities is obtained by taking out pixels whose densities are not similar to the density qi of the reference pixel bi in each calculation area Bi and obtaining the average value. Therefore, the average value of the dissimilar densities takes a value that is inverted with respect to the density qi of the reference pixel bi. The calculation area Bi set when calculating the dissimilar density average value is preferably larger than the calculation area Ai set when calculating the moving average value of density. When the calculation area Bi is small, the possibility that a pixel having a density that is not similar to the reference pixel bi exists in the calculation area Bi is reduced. In order to obtain a pixel whose density is not similar to the reference pixel bi, the calculation area Bi needs to be relatively large. On the other hand, if the calculation area Bi is too large, the density gradient or tendency of the image data for which the dissimilar density average is obtained is diluted.

  In the method of obtaining the moving average of the dissimilar density in this example, image data that is the target for obtaining the moving average of the dissimilar density is acquired in step S401. This image data is the image data smoothed in step S202 of FIG. That is, the image data indicated by the curve 7021 in the graph 702 of FIG. However, the image data for which the moving average of the dissimilar density is obtained may be the original image data acquired in step S201 in FIG. That is, it may be the original image data indicated by the curve 7011 of the graph 701 in FIG.

  Further, in step S404, a predetermined number of pixels having a large difference from the density qi of the reference pixel bi among the pixels included in each calculation region Bi are extracted in order of increasing density difference. The reference pixel bi is extracted from the smoothed image data indicated by the curve 7021 of the graph 702 in FIG. However, for each calculation area Bi, a predetermined number of pixels having a large difference from the density pi of the reference pixel ai of the original image data may be extracted from among the pixels included in the calculation area Bi. That is, a pixel having a large density difference as compared with the reference pixel ai extracted from the original image data indicated by the curve 7011 of the graph 701 in FIG. 7 may be extracted.

  If the original image data before obtaining the moving average value is acquired in step S401 and the reference pixel obtained from the original image data before obtaining the moving average value in step S404 is used, the process proceeds to step S202. The process of calculating the moving average of density is not necessary. Therefore, according to the present invention, the process of step S202 may be omitted in the noise removal process of FIG.

  Although the examples of the present invention have been described above, the present invention is not limited to the above-described examples, and it is easily understood by those skilled in the art that various modifications can be made to the invention described in the claims. Like.

  The noise removal apparatus and method of the present invention can be applied to a processing for detecting a coating defect in a painting inspection apparatus for an automobile body, but can also be applied to noise removal techniques in other technical fields.

It is a figure which shows the outline of the coating inspection apparatus of the vehicle body of the motor vehicle by this invention. It is a figure which shows the outline of the detection process of the defect of the coating in the coating inspection apparatus of the vehicle body of the motor vehicle by this invention. It is a figure which shows the outline of the noise removal process by this invention. It is a figure which shows the outline of the calculation process of the moving average of the density | concentration in the noise removal process by this invention. It is a figure which shows the outline of the calculation process of the moving average of dissimilar density in the noise removal process by this invention. It is a figure which shows the example of the original image used as the object of the noise removal process by this invention, and the image after a noise removal process. It is a figure which shows the example of the density distribution of the image data in the noise removal process by this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 ... Car body, 11 ... Illuminating device, 12 ... Lens unit, 13 ... Camera, 14 ... Image processing device, 601, 602 ... Image, 6012 ... Noise, 6013, 6023 ... Defect of painting

Claims (7)

In a method for removing noise from image data using an image processing device having an arithmetic processing device and a storage device,
An image data acquisition step of acquiring image data including the density of each pixel obtained by the imaging device by the arithmetic processing device;
A non-similar density average calculating step of calculating a moving average of non-similar density from the density of each pixel by the arithmetic processing unit;
Subtracting the moving average of the dissimilar density from the density for each pixel by the arithmetic processing unit, and obtaining a difference data calculation step for obtaining a difference between the two,
The dissimilar density calculation step includes:
A reference pixel acquisition step in which pixels are sequentially extracted from the pixels constituting the image data at a predetermined pitch, and are set as reference pixels bi (i = 1 to k).
For each of the reference pixels bi (i = 1 to k), a calculation area setting step for setting a predetermined area including the reference pixel and setting it as a calculation area Bi;
For each of the calculation areas Bi, non-similar density pixel acquisition that extracts a predetermined number of pixels having a large difference from the density qi of the reference pixel bi among the pixels included in the calculation area in order of increasing density difference Steps,
For each of the calculation areas Bi, calculating an average value cmi of the density of the extracted pixels;
Replacing each of the reference pixel densities qi with each of the average density cmi of the dissimilar densities;
The noise removal method characterized by including. In the noise removal method of Claim 1,
A density average calculation step further between the image data acquisition step and the dissimilar density average calculation step;
The concentration average calculation step includes:
A reference pixel acquisition step in which pixels are sequentially extracted from the pixels constituting the image data at a predetermined pitch, and are referred to as reference pixels ai (i = 1 to k).
For each of the reference pixels ai (i = 1 to k), a predetermined region including the reference pixel is set, and a calculation region setting step using the predetermined region as the calculation region Ai;
An average value calculating step for calculating an average value pmi of the density of pixels included in the calculation area for each of the calculation areas Ai;
Replacing each density pi of the reference pixel with each of the average values pmi of the densities;
And the calculation area Bi set by the calculation area setting step of the dissimilar density calculation step is larger than the calculation area Ai set by the calculation area setting step of the previous density average calculation step Removal method. In the noise removal method of Claim 2,
The noise removal method, wherein the image data used in the non-similar density average calculation step is image data after being processed in the density average calculation step. In the noise removal method of Claim 2,
For each of the calculation areas Bi in the dissimilar density pixel acquisition step, out of the pixels included in the calculation area, the density pi of the reference pixel ai acquired in the reference pixel acquisition step of the density average calculation step A noise removal method characterized by extracting a predetermined number of pixels having a large density difference in descending order of density difference. Illuminating the vehicle body of the vehicle with the illumination device; and
Capturing a digital image of the vehicle body with an imaging device;
Detecting a coating defect from an image of a vehicle body by an image processing apparatus having a memory, an arithmetic processing unit, and a storage device;
The step of detecting a defect in the coating includes
An image data acquisition step of acquiring image data including the density for each pixel from the imaging device by the arithmetic processing device;
A non-similar density average calculating step of calculating a moving average of non-similar density from the density of each pixel by the arithmetic processing unit;
Subtracting the moving average of the dissimilar density from the density for each pixel by the arithmetic processing unit, and obtaining a difference data calculation step for obtaining a difference between the two,
The dissimilar density calculation step includes:
A reference pixel acquisition step in which pixels are sequentially extracted from the pixels constituting the image data at a predetermined pitch, and are set as reference pixels bi (i = 1 to k).
For each of the reference pixels bi (i = 1 to k), a calculation area setting step for setting a predetermined area including the reference pixel and calculating the predetermined area;
For each of the calculation areas Bi, extracting a predetermined number of pixels having a large difference from the density qi of the reference pixel bi among the pixels included in the calculation area;
For each of the calculation areas Bi, calculating an average value cmi of the density of the extracted pixels;
Replacing each of the reference pixel densities qi with each of the dissimilar density averages cmi;
A coating inspection method characterized by comprising: In the coating inspection method according to claim 5,
A density average calculation step further between the image data acquisition step and the dissimilar density average calculation step;
The pre-concentration average calculation step is:
A reference pixel acquisition step in which pixels are sequentially extracted from the pixels constituting the image data at a predetermined pitch, and are referred to as reference pixels ai (i = 1 to k).
For each of the reference pixels ai (i = 1 to k), a calculation area setting step for setting a predetermined area including the reference pixel and calculating the predetermined area;
An average value calculating step for calculating an average value pmi of the density of pixels included in the calculation area for each of the calculation areas Ai;
Replacing each density pi of the reference pixel with each of the average values pmi of the densities;
And the calculation area Bi set by the calculation area setting step is larger than the calculation area Ai set by the calculation area setting step. An illumination device that irradiates a vehicle body with illumination light, an optical system that enlarges or reduces an image of the vehicle body, an imaging device that captures a digital image of the vehicle, and image processing that detects a paint defect from the image of the vehicle And an image processing apparatus comprising: a storage device; an arithmetic processing device; and a coating inspection device including a memory.
The arithmetic processing unit acquires image data including the density for each pixel from the imaging device, calculates a moving average of the dissimilar density from the density for each pixel, and the moving average of the dissimilar density from the density for each pixel Is configured to obtain the difference between the two,
The moving average of the dissimilar concentrations is
From the pixels constituting the image data, one pixel is taken out in order at a predetermined pitch, and this is taken as a reference pixel bi (i = 1 to k),
For each of the reference pixels bi (i = 1 to k), a predetermined area including the reference pixel is set, and is calculated as a calculation area.
For each of the calculation areas Bi, out of the pixels included in the calculation area, a predetermined number of pixels having a large difference from the density qi of the reference pixel bi are extracted in descending order of density difference, and the calculation area Bi For each of the above, an average value cmi of the density of the extracted pixels is calculated,
A coating inspection apparatus, wherein each of the reference pixel densities qi is replaced by each of the dissimilar density average values cmi.

Images