JP2585872B2 - Image noise removal device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image noise removing device used in an image processing device. The image processing device is a device that can process visual information similar to humans.
It is actively applied to unmanned monitoring equipment and automatic guided vehicles. The potential demand is so high that research and development is even more active. In an image processing apparatus, the role of an image noise removing unit that removes image noise due to scattering of external light or the like is extremely large. This is because the original processing target can be effectively recognized for the first time by efficiently removing noise from the image.

[0002]

2. Description of the Related Art FIG. 7 is an explanatory view of the prior art, and FIG. 8 shows an example of noise removal according to the prior art. Some image processing apparatuses receive a binary image, and others receive a grayscale image (including a color grayscale image). To simplify the explanation,
An image processing apparatus that receives a binary image as an input will be described as an example. The vicinity of the pixel of interest is a 3 × 3 pixel area. Further, the binary value is set to 0 (called white) or 1 (called black).

In a conventional image noise eliminator using a binary image as an input, for example, as shown in FIG.
For the × 3 pixel area, the state of the pixel 90 is determined, and a process of removing an isolated white point (value is 0) or a black point (value is 1) from the value of the central target pixel and the peripheral pixels is performed. Have been done. That is, as shown in (b) and (c) of FIG.
If all pixels obliquely above, below, left, and right of the pixel of interest are white points, the value of the pixel of interest is also updated to a white point. Update to

The configuration of a conventional image noise elimination device is as shown in FIG. Attention pixel value input unit 91
Inputs the value of the pixel of interest in the order of the scanning lines. The peripheral pixel value input unit 92 inputs the values of pixels obliquely above, below, left, and right of the target pixel in the order of scanning lines. The AND circuit 93 calculates the AND value of the pixel value input to the peripheral pixel value input unit 92. O
The R circuit 94 calculates the OR value of the pixel value input to the peripheral pixel value input unit 92. The target pixel value determination output unit 95
When the output value of the AND circuit 93 is 1, 1 is output and O
When the output value of the R circuit 94 is 0, 0 is output. Otherwise, the output value of the target pixel value input unit 91 is output.

[0005] In a conventional image noise eliminator using a gray-scale image as input, a process of replacing an isolated pixel value with an average value of neighboring pixel values is performed. In a conventional image noise elimination device that receives a color image as an input, the above-described conventional method is applied according to whether each of the RGB signal lines is a binary image or a grayscale image.

[0006]

The conventional image noise eliminator using a gray-scale image as input has the disadvantage that only isolated noise can be eliminated. The general image screen is 640 × 4
It is composed of many pixels, such as 80 pixels or 512 × 512 pixels, and has many noises other than isolated points. For example, there are two isolated noises.

For example, it is assumed that image data as shown in FIG. The original image is only vertical lines, but there is one or two isolated noises. The conventional image noise elimination device can eliminate or remove noise at an isolated point, but cannot eliminate the noise. Therefore, the processing result by the conventional image noise elimination device is as shown in FIG. SUMMARY OF THE INVENTION It is an object of the present invention to solve the above problems and to provide a means for efficiently removing image noise.

[0008]

FIG. 1 shows a configuration example of the invention according to the first aspect. The image noise elimination device shown in FIG.
Noise is removed by inputting image data of a grayscale image. For this reason, the pixel-of-interest neighboring value input unit 20 for inputting the pixel value in the region near the pixel of interest and the pixel in the neighborhood of the pixel of interest having a value equal to or less than a certain lower limit are equal to or greater than a preset number. Updates the value of the pixel of interest to the average of the values of the pixels below the lower limit, and if the number of pixels having a value above a certain upper limit is greater than or equal to a preset number, the value of the pixel of interest is A target pixel value change output unit 21 that updates the average value of the pixels equal to or larger than the value and does not change the value of the target pixel in other cases is provided. Note that the pixel of interest may or may not be included as a pixel value in a region near the pixel of interest. The same applies to the following inventions.

A second aspect of the present invention relates to an image noise removing apparatus which receives a color RGB grayscale image as an input. The apparatus according to the present invention includes a target pixel neighboring value input unit 20 and a target pixel value change output unit 21 shown in FIG. 1 for each of the RGB signals.

FIG. 2 shows a configuration example of the third aspect of the present invention. The configurations of the target pixel neighboring value input unit 30 and the target pixel value change output unit 31 are the same as those in FIG. In the present invention,
An image temporary storage unit 39 for temporarily storing the output value of the target pixel value change output unit 31 is provided. Then, for the pixels that have already been processed by the target pixel value change output unit 31 as the neighboring values of the target pixel, the updated pixel values are used by the input from the image temporary storage unit 39.

According to the fourth aspect of the present invention, FIG.
The image noise elimination device shown in FIG.

FIG. 3 shows a configuration example of the invention according to claim 5. The target pixel neighboring value input unit 40 has an address conversion unit 42 that generates an address from a neighboring value of the target pixel. The target pixel value change output unit 41 includes a memory 43 that stores an output value corresponding to a change condition of the target pixel in advance at a position corresponding to an address generated from a nearby value of the target pixel. By referring to the memory 43 by the address generated by the unit 40, the value of the pixel of interest is changed, and noise is removed.

[0013]

The present invention is an extension of the conventional idea of removing isolated points in order to remove not only noise at one isolated point but also general noise. In particular, each pixel is represented by multi-values. This makes it possible to remove noise from a grayscale image or a color RGB grayscale image.

In the image noise elimination device for inputting a gray-scale image according to the present invention, when the number of pixels having a value equal to or less than a certain lower limit in a region near the pixel of interest is equal to or greater than a predetermined number, the value of the pixel of interest is determined. Is updated to the average of the values of pixels below the lower limit, and if the number of pixels with values above a certain upper limit is more than a preset number, the value of the pixel of interest is The value is updated to the average value, and otherwise, the value of the pixel of interest is not changed. If the threshold value of the number of pixels at this time can be set from outside,
It is also possible to appropriately remove noise according to various environments.

Further, in an image noise elimination device which receives a color RGB grayscale image as input, noise elimination processing is performed on the grayscale image for each of the RGB signals.

As for the vicinity value of the pixel of interest, for a pixel for which noise removal processing has already been performed, this may be fed back and the updated value of the pixel may be used. Further, by combining the above-described image noise elimination devices in multiple stages, the effect of noise elimination can be further enhanced.

Rather than determining the pixel value in the vicinity of the pixel of interest each time to remove noise, the value of the pixel of interest is determined by a memory table search method using an address generated from the vicinity of the pixel of interest. It may be determined.

[0018]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS First, the configuration example of FIG. 1 will be described. The target pixel neighboring value input unit 20 inputs a pixel value near the target pixel. The input is, for example, a value near a 3 × 3 pixel. The lower limit storage unit 22 stores “lower limit”. The below-lower-limit value counting unit 23 is a unit for inputting the neighborhood value of the pixel of interest.
Among the pixel values stored in 0, the value of the pixel below the “lower limit” and the number of pixels are counted.

The upper limit value storage section 24 stores an "upper limit value". The upper limit value or more counting unit 25 counts the value of the pixel value equal to or more than the “upper limit value” and the number of pixels among the pixel values stored in the target pixel neighboring value input unit 20. Lower limit average number storage unit 26
Stores the “lower limit average number”. The upper limit average number storage unit 27 stores “upper limit average number”.

When the number of pixels counted by the count unit 23 below the lower limit value is larger than the “lower limit average number”, the target pixel value determination output unit 28 outputs the average value of the pixel values equal to or less than the “lower limit value”. I do. If the number of pixels counted by the counting unit 25 is equal to or greater than the upper limit value, the average value of the pixel values equal to or greater than the “upper limit value” is output. In other cases, the value of the target pixel is output as it is.

FIG. 4 shows the result of processing by the image noise elimination device according to the embodiment of the present invention. For the input image shown in FIG. 4A, both the lower limit average number stored in the lower limit average number storage unit 26 and the upper limit average number stored in the upper limit average number storage unit 27 in FIG. Processing. In the figure, white circles indicate pixels whose pixel values are equal to or lower than the lower limit, black circles indicate pixels whose pixel values are equal to or higher than the upper limit, and circles marked with x indicate pixels considered to be noise. When noise removal processing is performed on this input image element, the result is as shown in FIG.
It can be seen that the noise has been clearly removed.

FIG. 5 (a) shows an example of an input image having more noise than the image shown in FIG. 4 (a).
According to the embodiment of the present invention, when the lower limit average number is set to “5” and the upper limit average number is set to “5” to remove noise, the processing result is as shown in FIG.
(B). It can be seen that the noise is almost completely removed. Further, noise can be further removed by combining the image noise removing apparatus of the present invention in multiple stages.

The configuration example of the present invention shown in FIG. 2 is as follows. In this embodiment, as the neighborhood value of the pixel of interest,
For the pixels that have already been processed, updated pixel values are used. Points different from the configuration example shown in FIG. 1 will be described. The image temporary storage unit 39 temporarily stores one image screen. The target pixel neighboring value input unit 30 sequentially inputs pixel values near the target pixel from the image temporary storage unit 39. The target pixel value determination output unit 38 outputs the determined value of the target pixel to the outside and stores it in the image temporary storage unit 39.

As described above, by feeding back the noise removal result to the original image, it is possible to obtain an effect of further removing noise.

In the configuration example of the present invention shown in FIG. 3, a memory is used to obtain an output value from which noise has been removed. For example, if the neighborhood value of a 3 × 3 pixel is determined, the value of the pixel of interest should be determined. Therefore, the output value setting unit 46 calculates output values in advance for all possible values of the pixels in the vicinity of the target pixel, and generates the obtained output values from the neighboring values of the target pixel in the memory 43. Stored at the address location.

In actual use, in the target pixel neighborhood value input unit 40, the address conversion unit 42 converts the neighborhood value of the target pixel by image data input into an address, and reads the memory 43 to obtain the address shown in FIG. The result equivalent to the image noise elimination device shown in FIG.

FIG. 6 shows the configuration example shown in FIG. 3 more specifically. Hereinafter, description will be made in accordance with FIG. For simplicity, the case of a binary image will be described as an example, but application to multi-valued (shaded images) can be easily analogized from the following description. In the application to a grayscale image, A0 to A8 in the following description each have several bits.

The address converter 42 converts image data input flowing in real time into address information according to a certain rule. The microprocessor (MPU) unit 70
A selection signal line is output to the to1 selector 71.
The 2to1 selector 71 selects one of the address obtained by the address conversion unit 42 and the address output from the MPU unit 70 and outputs the selected address to the memory 43.

When the address obtained by the address conversion section 42 is selected, the memory 43 is read at that address and the numerical value is output as image data. When the address of the MPU unit 70 is selected, the memory 43
The data is read / written according to the control of the read / write / data lines from the PU unit 70.

When actually functioning as an image noise elimination device, first, the 2to1 selector 71
The address of the unit 70 is selected, and appropriate data is written to the memory 43. Thereafter, in the 2to1 selector 71, the address data from the address conversion unit 42 is selected, and the image data from which noise has been removed according to the input image data is output from the memory 43.

FIG. 6B shows an embodiment of the address conversion unit 42. In this embodiment, 3
A region of × 3 is taken. Line buffers 73 and 74
Is a buffer for one line of the image, and outputs the image one line behind the input image. A0 to A8 are registers that output an image with a delay of one pixel from the input image. Thus, the address conversion unit 42
A0 to A8 are output as address information.

FIG. 6C shows the correspondence between registers A0 to A8 and pixels. ◎ is a target pixel, and moves to the right by one pixel for each process. A0 has the value of the target pixel, and A1 to A8 have the values of neighboring pixels.

In this example, the correspondence between the registers A0 to A8 and the addresses of the memory is as shown in FIG. A0 corresponds to the most significant bit (MSB),
8 corresponds to the least significant bit (LSB). In the case of a grayscale image (multi-valued image), each of A0 to A8 is composed of several bits.

By combining image noise eliminators as shown in FIG. 6A in multiple stages, more powerful noise elimination can be performed. At this time, the MPU unit 70 can obviously be shared.

The embodiment of the present invention for a monochrome grayscale image has been described above. However, the present invention is also applicable to an image noise elimination device which receives a color RGB grayscale image as input for each monochrome RGB grayscale image. It can be applied by applying the embodiment.

[0036]

As described above, according to the present invention,
Various noises contained in the grayscale image or the color RGB grayscale image can be efficiently removed, and a clear image can be obtained. By using it for an image processing device,
It is possible to improve the recognition accuracy of an object or the like.

[Brief description of the drawings]

FIG. 1 is a diagram showing a configuration example of the present invention.

FIG. 2 is a diagram showing a configuration example of the present invention.

FIG. 3 is a diagram showing a configuration example of the present invention.

FIG. 4 is a diagram showing an example of noise removal according to the present invention.

FIG. 5 is a diagram showing an example of noise removal according to the present invention.

FIG. 6 is a diagram showing an embodiment of the present invention.

FIG. 7 is an explanatory diagram of a conventional technique.

FIG. 8 is a diagram showing an example of noise removal according to the related art.

[Explanation of symbols]

 20 attention pixel value neighborhood value input section 21 attention pixel value change output section 22 lower limit value storage section 23 lower limit value or less counting section 24 upper limit value storage section 25 upper limit value or less counting section 26 lower limit average number storage section 27 upper limit average number storage section 28 Attention pixel value judgment output section

Claims (5)

(57) [Claims] An image noise elimination device which receives a grayscale image as input, comprising: a pixel-of-interest neighboring value input means (20) for inputting a pixel value in a region near the pixel of interest; If the number of pixels having a value of is equal to or greater than a preset number, the value of the pixel of interest is updated to the average of the values of the pixels equal to or less than the lower limit, and the number of pixels having a value equal to or greater than a certain upper limit becomes If the number of pixels is equal to or more than a preset number, the value of the pixel of interest is updated to the average value of the pixels equal to or larger than the upper limit, otherwise, the value of the pixel of interest change output unit (21) which does not change the value of the pixel of interest ). 2. An image noise elimination device which receives a color RGB grayscale image as input, wherein, for each of the RGB signals, a target pixel neighboring value input means (20) according to claim 1, and a target pixel value according to claim 1 An image noise elimination device comprising a change output means (21). 3. The image noise elimination device according to claim 1, further comprising: an image temporary storage unit (36) for temporarily storing an output value of said target pixel value change output unit; As for the value of the pixel which has already been processed by the target pixel value change output means, the updated pixel value is used by the input from the image temporary storage means (36). Noise removal device. 4. An image noise eliminator characterized in that the image noise eliminator according to claim 1, 2 or 3 is combined in multiple stages. 5. The image noise elimination apparatus according to claim 1, wherein said target pixel neighboring value input means generates an address from a neighboring value of the target pixel. The target pixel value change output means has a memory (4) for storing in advance an output value corresponding to a change condition of the target pixel at a position corresponding to an address generated from a neighboring value of the target pixel.
An image noise eliminator configured by (3), wherein the value of the pixel of interest is changed by referring to the memory (43) according to the address generated by the pixel of interest neighboring value input means.