JPS6329889A - Image processing method - Google Patents

Abstract

PURPOSE:To almost completely remove noise existing in an image and to obtain an accurate image even in the succeeding processed result by successively driving noise removing operators for removing projecting parts existing in an extracted image. CONSTITUTION:An image processor is provided with a contour extracting part 6 for binarizing a digital signal stored in an image memory 5 and extracting the contour of an object by a primary differential or Laplacian operator, an image memory 7 for storing a binarized image, a filter part 8 for removing noise existing in the binarized image, and an output part 9 for outputting data relating to the noise-removed image. Although noise due to a variation component of the illumination of the object is usually included in the image obtained by the extracting part 6 at the time of input of the image, the noise is completely removed by the filter part 8 and the filtered image is stored in the image memory 7 and outputted from the output part 9 when necessary.

Description

DETAILED DESCRIPTION OF THE INVENTION (Industrial Application Field) The present invention provides, for example, an image input from an image input device such as a camera. The present invention relates to an image processing method that selectively removes only noise.

(Prior art) In recent years, with the amazing development of computers, in the industrial world, for example, cameras are connected to industrial robots to provide vision similar to that of humans, allowing them to understand the shape and position of workpieces, etc. Industrial robots equipped with so-called artificial intelligence have come into use, allowing them to measure and perform processing operations with greater flexibility.

By the way, as a method for processing visual images input from a camera connected to an industrial robot, binary image processing is widely used in practical visual systems.

This binarized image processing is, as shown in FIG. 5(a),
A workpiece 2 is imaged by a camera 1, and this imaged image is input to a computer (not shown). The computer converts this manually generated image into MXN as shown in Figure (b).
Each pixel is decomposed into 3 pixels, and each pixel is binarized using the P-tile method, mode method, etc., to create a binarized image as shown in FIG. This image generally contains noise 3° as shown in the same figure, for example, and such noise 3° is removed by the isolated point removal operator shown in (e) of the same figure and the noise (f) shown in the same figure. ) is removed by the degeneracy operator shown in ). Specifically, when the isolated point removal operator is applied to the image shown in FIG. Calculate the HAND between the matrix elements and the image up to, and as a result of the calculation,
If at least one matrix element has a value of 1, the central pixel is forcibly set to ○.

), it is possible to obtain an image as shown in FIG.

As another example, when a binarized image as shown in FIG. 6(a) is processed by the isolated point removal operator and degeneracy operator described above, images as shown in FIG. 6(b) and FIG. 6(C) are obtained. is obtained.

In this way, the isolated point removal operator purely removes only the isolated points that exist within the image, and the degeneration operator removes noise while removing part of the image.

(Problems to be Solved by the Invention) However, in such conventional image processing devices, noise existing in images may not be completely removed.
Alternatively, even if the line segment is removed, if the area of the image in the image and the area of the image in the processed image are significantly different, as shown in Figure 6, the thin line segment may be completely erased. This poses a problem in that large errors occur in subsequent processing results.

The present invention has been made in view of these conventional problems, and it is an object of the present invention to provide an image processing method that can effectively remove only the noise present in a binarized image. purpose.

(Means for Solving the Problems) In order to achieve the above object, the present invention decomposes an image captured by an image input device into a large number of pixels, sets a pixel that is considered to have an image to 1, and A binarized image is created by performing binarization processing on all the decomposed pixels, in which pixels that are considered to have no image are set to O, and nine pixels in 3 rows and 3 columns are extracted from the binarized image. Extract an arbitrary region consisting of 3 rows and 3 columns of matrix elements, including the three matrix elements in the outermost row or column, and among these matrix elements, the column direction or row for the corner matrix element. Eight types of noise removal operators having an arrangement in which one matrix element adjacent in the direction must be O are sequentially applied to the region, and O in the region is
When the pixel arrangement set to matches any of the eight types of noise removal operators having the above-mentioned O arrangement, the central pixels constituting the area are forcibly set to O. The present invention is characterized in that noise present in the binarized image is removed by performing this processing on all regions including overlapping regions of the decomposed pixels.

(Operation) When an image after binarization processing is processed using the above method, only the protruding parts of the image are removed by the noise removal operator, so even if the image is a thin line segment, it will be erased. Therefore, ideal noise removal can be performed.

(Example) Below, an example of the present invention will be described in detail based on the drawings.

FIG. 1 shows a general diagram of a general image processing apparatus. As shown in the figure, this image processing device includes a television camera 1 as an image input device that captures a pupil image of a subject, and is connected to the television camera 1 and converts an analog signal outputted from the television camera 1 into a digital signal. An A/D converter 4 is provided to store the digital signal, and an image memory 5 is provided to store the digital signal.The image input by the television camera 1 is sampled and quantized and stored in the image memory 5. .

Roughly speaking, the image@processing device includes a contour extraction section 6 that binarizes the digital signal stored in the image memory 5 and extracts the contour of the subject using an operator such as a first-order differential or Laplacian; An image memory 7 that stores the image after the digitization process, a filter part 8 that removes noise present in the image after the binarization process, and an output section 9 that outputs data regarding the image after the noise removal process are provided. The image obtained by the contour extraction section 6 usually contains noise due to fluctuation components of the subject illuminance at the time of image input, and this noise is completely removed by the filter section 8. The image after this processing is stored in the image memory 7, and is outputted from the output unit 9 as required.

Next, a noise removal operator according to the invention stored in the filter part 8 is shown in FIG.

As shown in the figure, there are eight types of noise removal operators. In the figure, a matrix element represented by * means that the element may be either 1 or O.

FIG. 3(a) schematically shows the functions of the calculations in the filter portion 8 described above. Furthermore, the matrices shown in FIGS. 13(b) and 2(c) show an arbitrary 3 rows and 3 columns area and a noise removal operator after the binarization process.

Next, the double (overlapping) method as shown in FIG. 4(a) is processed as follows using the filter part shown in FIG. 3 and the noise removal operator in FIG. This process will be explained in detail.

First, the filter part 8 extracts a region 10 consisting of nine pixels arranged in 3 rows and 3 columns as shown in FIG. 4(a), for example.
This territory! The noise removal operator shown in FIG. 2 is applied to 10. In other words, if area 10 is represented by digital values of ○ and 1, since there is no image within area 10, the digital 1 value of each pixel that makes up this area will be O, and this will be expressed as matrix A1. Next, the aforementioned eight types of noise removal operators N are applied to 1 on this matrix. For example, in the noise removal operator matrix A1 shown in FIG. 2(a), Since the configurations of the matrix elements are the same, the pixel at the center of region 10 is set to O. Similarly, consider the case where the region 11 shown in the figure is extracted, and if the extracted region 11 is represented by a matrix A2, then to this matrix 2,
Even if the eight types of noise removal operators N described above are applied sequentially as described above, the configuration of the matrix element that is O in matrix A2 will not match any of the noise removal operators N (,:). Therefore, in this case, the central pixel of 611i!j, 11 is not set to O. However, it is determined that there is no noise in this area.

Roughly, consider the case where the region 12 shown in the figure is extracted, and the extracted region 12 is represented by a matrix △3. Then, in this matrix A2, the eight types of noise removal operators N described above are written #J. When applied sequentially as shown in FIG. 2(C) and (g>), since the configurations of the matrix elements that are O in the matrices shown in FIG. In this case, the pixel 13 within the region 12 will be considered as noise and will be removed.

The above process can be expressed numerically as follows.

First, the coefficients of each matrix element shown in FIG. 3(C) are NANDed with W11='l and NANDed with W12=1.
NAND with OW13=1, NAN with W21=1
DW22=O, W23=O, W31=O If W32=O, W33=O, then Y=ΣWij
-X ij+ b l凰1 0"' (b is the bias value) When Y=0, set the value of the center pixel to O. When Y≠○, leave the value of the center pixel as it was. .

It will be done.

As described above, from among the pixels constituting the image shown in FIG. By sequentially applying the noise removal operators, the final processed image has the noise portion completely removed as shown in FIG.

(Effects of the Invention) As is clear from the above explanation, according to the present invention, the noise removal operator that removes the protrusions present in the extracted image is sequentially operated, so that Noise can be almost completely removed, and subsequent processing results can be improved in accuracy.

[Brief explanation of drawings]

FIG. 1 shows an overview of an image processing apparatus that performs image processing using the image processing method of the present invention. FIG. 2 is an 8-channel diagram, and FIG. 4(a) is a diagram schematically representing the calculation function of a part of the filter in the image processing apparatus according to the present invention. (b) is a diagram providing a concrete explanation of image processing by the image processing method according to the present invention, and FIGS. 5(a) to 6(f) to 6(a) to (C) are diagrams showing conventional FIG. 3 is a diagram illustrating a specific explanation of image processing using a processing method. DESCRIPTION OF SYMBOLS 1... Television camera (image input device), 2... Workpiece, 3... Pixel, 8... Filter part, 10.
11.12...Extraction area. Agent Patent Attorney 8 1) Mikio (and 2 others) 1st
Figure @2 Figure (d) (e) (f) ((
J) (h) Figure 3 Figure 4 (b) Figure 5 (a) (b) ft, +

Claims (1)

[Claims] Decomposes a binarization process that decomposes an image captured by an image input device into a large number of pixels, sets pixels that are considered to have an image to 1, and sets pixels that are considered to have no image to 0. Create a binarized image by extracting an arbitrary region consisting of 9 pixels arranged in 3 rows and 3 columns from the binarized image, and extract a region consisting of matrix elements arranged in 3 rows and 3 columns Three matrix elements in the outer row or column, and one of these matrix elements that is adjacent to the corner matrix element in the column or row direction.
8 types of noise removal operators having an array in which 1 matrix element and 0 are 0 are sequentially applied to the region,
When the arrangement of pixels set to 0 in the region matches any of the eight types of noise removal operators having the above-mentioned arrangement of 0, the central pixels forming the region are forced. Image processing characterized in that noise present in the binarized image is removed by performing a process of setting the pixels to 0 on all regions including overlapping regions of the decomposed pixels. Method.