JP2005252451A - Image quality inspection method and image quality inspection apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image quality inspection method capable of inspecting the propriety of an entire image with high accuracy. <P>SOLUTION: The image quality inspection method for carrying out image quality inspection of images obtained by image input apparatuses determines a Maharanobis reference space by using a luminance distribution extracted from image data of a good item collected by an image input apparatus of a good item and thereafter calculates a Maharanobis distance on the basis of a luminance distribution of an image obtained from an image input apparatus being an inspection object, and compares the Maharanobis distance with a prescribed threshold to discriminate the propriety of the image input apparatus being the object of inspection. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to an image quality inspection method and an image quality inspection apparatus. More specifically, the present invention relates to an image quality inspection method and an image quality inspection apparatus that attempt to improve the accuracy of image quality inspection by using the Mahalanobis distance calculated based on a predetermined feature amount distribution.

  Conventionally, the color of a color image, image blur (resolution), and the like are evaluated by different methods, and the quality of the color image is determined separately. Therefore, the quality of the overall color image quality is evaluated by human comparison by visual comparison with a sample of limit samples, or by a skilled evaluator performing sensory evaluation.

  However, in human visual evaluation, for example, even if there is a comparative sample such as a limit sample, there are individual differences in determination, and there is a problem in reproducibility. There is a problem that the efficiency is very poor because it cannot be judged.

  In order to solve the problems as described above, a predetermined feature amount extracted from predetermined image data in advance (for example, a color or luminance of a predetermined position or a predetermined region (luminance is the luminance of each primary color in the color space to be used). In the case of RGB 8-bit data, each of R, G, and B takes a value of 0 to 255.) Primary parameters such as density, secondary values such as average value and standard deviation of primary parameters, etc. Parameter) is used to calculate the Mahalanobis distance using a predetermined feature amount extracted from the inspection target image, and the calculated Mahalanobis distance is compared with a predetermined threshold value to determine the image of the inspection target image. An image quality inspection method for determining pass / fail has been proposed (see, for example, Patent Document 1 and Patent Document 2).

The Mahalanobis reference space is a reference space set from image data of only good products. In the field of quality optics, Dr. Genichi Taguchi developed using the distance devised by Indian statistician Mahalanobis. Defined in the Mahalanobis-Taguchi system (MTS method).
The Mahalanobis distance is a scale indicating the degree of abnormality of a sample to be evaluated, that is, how much the sample is deviated from normal data.

JP 2002-267533 A

JP 2000-275757 A

However, in a conventional image quality inspection method using, for example, primary parameters such as color, brightness, and density of a predetermined position or a predetermined region, and secondary parameters such as an average value and standard deviation of primary parameters as a predetermined feature amount. When the overall image information is insufficient and the quality of the entire image is determined, the inspection accuracy may decrease.
That is, in the primary parameter and the secondary parameter described above, it is difficult to express all of the images to be inspected because there are positions where the parameters are not measured. Further, when statistical parameters such as an average value and a standard deviation are used, it is still difficult to express all the images to be inspected although the information is in an area having a predetermined range. Accordingly, when determining the quality of the entire image, the inspection accuracy may be lowered.

  Although it is considered that the inspection accuracy can be improved by increasing the number of parameters, in order to determine the Mahalanobis reference space for calculating the Mahalanobis distance, it is generally more than twice the number of parameters. Since samples are needed, the number of parameters cannot be increased unnecessarily.

  The present invention has been made in view of the above points, and an object of the present invention is to provide an image quality inspection method and an image quality inspection apparatus that can accurately inspect the quality of an entire image.

  In order to achieve the above object, an image quality inspection method according to the present invention includes a step of defining a Mahalanobis reference space using a predetermined feature amount distribution extracted from image data in an image quality inspection method for performing image quality inspection of an image, A step of calculating a Mahalanobis distance in the Mahalanobis reference space based on a predetermined feature amount distribution extracted from image data of the inspection target image and comparing the Mahalanobis distance with a predetermined threshold value to determine whether the inspection target image is good or bad The process of carrying out is provided.

  In order to achieve the above object, an image quality inspection apparatus according to the present invention extracts image data input means for inputting image data, and a predetermined feature amount distribution from the image data input by the image data input means. A Mahalanobis reference space setting means for defining a Mahalanobis reference space using the extracted predetermined feature quantity distribution, and extracting a predetermined feature quantity distribution from the image data of the inspection target image input by the image data input means, and extracting A Mahalanobis distance calculating means for calculating a Mahalanobis distance in the Mahalanobis reference space based on the predetermined feature amount distribution, and a determining means for comparing the Mahalanobis distance with a predetermined threshold to determine whether the inspection target image is good or bad. .

Here, by determining the Mahalanobis reference space using a predetermined feature amount distribution extracted from the image data, the Mahalanobis reference space can be determined based on a parameter including information on the entire image.
Further, by calculating the Mahalanobis distance in the Mahalanobis reference space based on a predetermined feature amount distribution extracted from the image data of the inspection target image, the Mahalanobis distance in the Mahalanobis reference space is calculated based on a parameter including information on the entire image. Can be calculated.

  In the above-described image quality inspection method and image quality inspection apparatus of the present invention, the predetermined feature amount is not used as a parameter, but a predetermined feature amount distribution is used as a parameter. Can be accurately determined.

Hereinafter, embodiments of the present invention will be described with reference to the drawings to facilitate understanding of the present invention.
FIG. 1 is a block diagram for explaining an example of an image quality inspection apparatus to which the present invention is applied. The image quality inspection apparatus 1 shown here is a solid-state imaging device, digital camera, camcorder, camera module for converting an image into electronic data. The image input device 2 such as a scanner, the image processing means 3 for calculating the luminance distribution from the luminance information of the pixels in the evaluation target area based on the electronic data input from the image input device, and the image determined to be non-defective Mahalanobis reference space (database) 4 determined from the luminance distribution in the evaluation target area of the electronic data, Mahalanobis distance calculation means 5 for calculating the Mahalanobis distance in the Mahalanobis reference space from the luminance distribution of the electronic data to be inspected, and calculation An image quality evaluation means 6 for evaluating the image quality by comparing the Mahalanobis distance with a predetermined threshold value. It is.

  Hereinafter, an image quality inspection method using the above-described image quality inspection apparatus will be described. That is, an example of an image quality inspection method to which the present invention is applied will be described.

  In an example of an image quality inspection method to which the present invention is applied, first, an image sample serving as a reference for evaluation is collected in order to set a Mahalanobis reference space. Here, the Mahalanobis reference space serves as a reference for evaluation of image anomalies, and it is necessary to collect normal data to be used as a reference for evaluation.

  In this embodiment, in order to inspect the drawing ability of the image input device 2, an electronic image of 640 × 480 pixels is collected as data using a general chart by a non-defective image input device, and the image processing means 3 From the luminance (8 bits: 0 to 255) data of red, green, and blue at each pixel of 640 × 480 pixels, 16 of 0 to 15, 16 to 31, 32 to 47,..., 240 to 255, respectively. The number of pixels in the luminance range of the stage was counted and used as luminance distribution data. Similarly, luminance distribution data is calculated for 120 non-defective image input devices, and the Mahalanobis reference space 4 is set.

  FIG. 2 shows the result of calculating the Mahalanobis distance (MD) of each good sample in the Mahalanobis reference space obtained as described above. As can be seen from FIG. 2, the Mahalanobis distance is less than 3 for a good sample.

  Next, an electronic image of 640 × 480 pixels is collected as electronic data using a general chart by the image input device of (1) to (13) to be inspected, and each of 640 × 480 pixels is collected by the image processing means. Pixels in the luminance range of 16 levels from 0 to 15, 16 to 31, 32 to 47,..., 240 to 255 from the luminance data (8 bits: 0 to 255) of red, green, and blue, respectively. The number is counted, the luminance distribution data of the image obtained by the image input device to be inspected is obtained, and the Mahalanobis distance is calculated by the Mahalanobis distance calculation means 5 based on the Mahalanobis reference space described above. Of the image input devices (1) to (13) to be inspected here, the image input devices (1) to (8) are defective and the image input devices (9) to (13). Is a good product.

  Table 1 shows the Mahalanobis distance of the image input devices (1) to (13) obtained as described above.

  From Table 1, the Mahalanobis distance of the defective image input devices (1) to (8) is a large value of 3.8 or more, and the non-defective product (9) to (13) Mahalanobis distance is the maximum. However, it was 1.5, and the Mahalanobis distance of the defective image input device was clearly different from the Mahalanobis distance of the non-defective image input device.

  In the case of the present embodiment, the non-defective sample from FIG. 2 has a Mahalanobis distance smaller than 3, and therefore MD = 3 as a threshold for determining that the image evaluation means 6 is defective. Set to.

  In the image quality inspection method and the image quality inspection apparatus to which the present invention is applied, since the calculation results of the Mahalanobis distance of the non-defective image input device and the Mahalanobis distance of the defective image input device are clearly different, each good product in the Mahalanobis reference space By setting an appropriate threshold value based on the Mahalanobis distance of the sample, it is possible to determine the quality of the image input device with high accuracy.

  In this embodiment, the image quality inspection of the image obtained by the image input device is performed and the case of evaluating the quality of the image input device is described as an example. However, the inspection target is limited to the image input device. The image quality of the image output from a display device such as a display can be checked to evaluate the quality of the display device.

  Here, for comparison, the image quality inspection of the image input devices (1) to (13) was performed by the conventional image quality inspection method as shown below. Similarly to the above, among the image input devices (1) to (13), the image input devices (1) to (8) are defective, and the image input devices (9) to (13) are It is a good product.

  That is, first, an electronic image of 640 × 480 pixels is collected as data using a general chart by a non-defective image input device, and an image of 640 × 430 pixels is divided into 12 blocks of 4 × 3. A Mahalanobis reference space was set with 500 non-defective samples using a total of 144 parameters such as the average value, standard deviation, maximum value, and minimum value of the luminance values of red, green, and blue. As a result of calculating the Mahalanobis distance of each good product sample, it was found that the Mahalanobis distance was less than 3 for the good product sample.

  Next, an electronic image of 640 × 480 pixels is collected as electronic data using a general chart by the image input device of (1) to (13) to be inspected, and an image of 640 × 480 pixels is 4 × 3. The Mahalanobis distance based on the Mahalanobis reference space described above using 144 parameters such as the average value, standard deviation, maximum value, and minimum value of the luminance values of red, green, and blue for each block. Was calculated.

  Table 2 shows the Mahalanobis distance of the image input devices (1) to (13) obtained as described above.

From Table 2, the Mahalanobis distance of the defective image input device (6) is 2.1, and the defective (7) image input device Mahalanobis distance is 1.8, both of which are less than 3. It is a value.
In other words, in the conventional image quality inspection method, there are those in which the Mahalanobis distance of the defective image input device is also less than 3, which makes it difficult to determine whether it is good or bad.

It is a block diagram for demonstrating an example of the image quality inspection apparatus to which this invention is applied. It is a graph which shows the result of having computed the Mahalanobis distance of each good product sample of Mahalanobis reference space.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Image quality inspection apparatus 2 Image input device 3 Image processing means 4 Mahalanobis reference space 5 Mahalanobis distance calculation means 6 Image quality evaluation means

Claims (4)

In an image quality inspection method for performing image quality inspection of an image,
A step of defining a Mahalanobis reference space using a predetermined feature amount distribution extracted from image data;
Calculating a Mahalanobis distance in the Mahalanobis reference space based on a predetermined feature amount distribution extracted from the image data of the inspection target image;
An image quality inspection method comprising a step of comparing the Mahalanobis distance with a predetermined threshold to determine whether the inspection target image is good or bad. The image quality inspection method according to claim 1, wherein the predetermined feature amount distribution is a luminance distribution. Image data input means for inputting image data;
A Mahalanobis reference space setting means for extracting a predetermined feature amount distribution from the image data input by the image data input means and defining a Mahalanobis reference space using the extracted predetermined feature amount distribution;
A Mahalanobis distance calculating unit that extracts a predetermined feature amount distribution from the image data of the inspection target image input by the image data input unit and calculates a Mahalanobis distance in the Mahalanobis reference space based on the extracted predetermined feature amount distribution. When,
An image quality inspection apparatus comprising: a determination unit that compares the Mahalanobis distance with a predetermined threshold to determine whether the inspection target image is good or bad. The image quality inspection apparatus according to claim 3, wherein the predetermined feature amount distribution is a luminance distribution.

Images