WO2001061989A1

WO2001061989A1 - Image processing method and image processing device

Info

Publication number: WO2001061989A1
Application number: PCT/JP2000/000978
Authority: WO
Inventors: Hiroshi Oshio
Original assignee: Fujitsu Limited
Priority date: 2000-02-21
Filing date: 2000-02-21
Publication date: 2001-08-23

Abstract

An image processing method and device for processing a magnified image, a restructured image restructured from a compressed image, and an artificial image into images imparting an impression of natural images. A continuous region in which the level difference is at a predetermined or lower level is detected in an image. It is judged whether or not noise should be added to the detected continuous region. Noise is generated and added to each image in the continuous region to which noise has been judged to be added. Noise is added according to statistic data to the continuous region where colors are similar and the levels are continuous, causing the image to be unnaturally uniform.

Description

Description Image processing method and image processing device

The present invention relates to an image processing method and an image processing apparatus for digital images, and more particularly to a case where low-resolution image data is printed or displayed at a high resolution, a case where a high-compression image is reproduced, or image data which has been subjected to an enlargement process. TECHNICAL FIELD The present invention relates to an image processing method and an image processing apparatus in a case where a certain area is printed or displayed with a uniform color and density, such as when printing or displaying an image. Background art

When printing or displaying a low-resolution image with a high-resolution image forming apparatus, it is necessary to convert the original low-resolution image into a high-resolution image in order to obtain an image of a desired size. For example, if 72-dpi image data with a size of 5 cm x 5 cm is supplied to a 600-dpi printer as it is, the size of the output image will be 1-8, that is, 0.6 mm 0. 6 mm. In this case, the image size of the original 5 cm x 5 cm and 72 dpi image is 14 2 pixels XI 42 pixels, and the size of 5 cm x 5 cm at 600 dpi is 1 It is 18 1 pixel x 1 18 1 pixel. Therefore, it is necessary to assign 8 pixels × 8 pixels to one original pixel.

There are various allocation methods. The simplest method is to assign 8 pixels × 8 pixels of the same color and level to one original pixel. Also, the pixel at the predetermined position of 8 pixels × 8 pixels is set to the same color and level as the original one pixel, and the other pixels are interpolated with the adjacent pixels of the predetermined position of 8 pixels × 8 pixels to smoothly interpolate. Those who want to change to There is also a law.

In the method of assigning 8 pixels x 8 pixels of the same color and level to one original pixel, 64 pixels of 8 pixels x 8 pixels are represented by the same color and level. Also, even in the method of interpolating and changing smoothly, if adjacent pixels have the same color and level in the original image, or if the difference is small, images of the same color and level may be continuous. As a result, a large area in which pixels of similar colors and levels are continuous can be formed.

A similar layout method is applied when enlarging the original image, and a large area where pixels of similar colors and levels are continuous may be formed.o

In addition, image data is compressed in order to transfer and store the image data. For example, in general JPEG, compression is performed in units of 8 pixels × 8 pixels in accordance with the level of the spatial frequency. When the spatial frequency is low, it is compressed into one piece of information, so when decompressed, 8 pixels x 8 pixels have the same color and level. Also, even if they are not a single, they are represented by a rectangular array of a few levels. In the original image, if there are slight differences between pixels, but similar colors and levels are widely continuous, the same color and level units will be continuous if they are grouped in units of 8 pixels x 8 pixels Cases arise. When such data is restored, a region in which pixels of the same color and level are continuous over a wide range occurs.

As described above, a region in which pixels of the same color and level are continuous over a wide range occurs even in an artificially created image such as a computer graphic.

Images like the ones described above are subject to different impressions than natural images taken with the camera. This is mainly due to gradients and drawing planes. This is due to the uniform or simple repetition of the change.

If you enlarge a photograph of an object painted with a single color using a camera, even in such a part, pixels of a single density are not continuous, and pixels of pixels with slightly different densities are not It is represented by a gathering. On the other hand, in the image as described above, a region where pixels of the same color and level are continuous over a wide range is formed.

Figure 1 is a diagram showing an example of the above image. Since the body of the train is painted with the same intention in each area, as shown in the enlarged part, only the color border is a small strip. Area is formed, but the other parts A and B are regarded as the same color, and a large continuous area is formed. In the case of a blue sky without clouds, the portion indicated by C is also determined to be almost a continuous region, and the portion D is determined to be a similar white continuous region.

In such a case, there was a problem of giving a very unnatural impression. Disclosure of the invention

The present invention solves such a problem, and performs processing such that the enlarged image, the restored image obtained by restoring the compressed image, and the artificially created image become images giving a more natural impression. The purpose is to realize an image processing method and apparatus.

In order to achieve the above object, according to the image processing method and apparatus of the present invention, noise (noise) is added only to a region where similar colors and levels are continuous.

That is, the image processing method of the present invention detects a continuous area having a level difference of a predetermined value or less in an image, determines whether to perform noise addition processing on the detected continuous area, and performs the noise addition processing. Noise is generated and added to each image in the continuous area. And

Further, the image processing apparatus of the present invention has an image data memory for holding image data, and a continuous area detection unit for detecting a continuous area having a level difference of a predetermined value or less in an image from the image data. An additional processing application determination unit that determines whether to perform noise addition processing on the continuous area, and a noise addition processing that generates and adds noise to each image in the continuous area on which noise addition processing is performed. According to the image processing method and apparatus of the present invention, the image processing apparatus according to the present invention is characterized in that the statistical processing is performed on unnaturally uniform continuous areas having similar colors and levels. The noise addition processing based on the day and night gives a natural impression. As the statistical data, for example, a normal distribution function or a uniform distribution function is used, and the standard deviation and the like are appropriately determined based on the average value of each color signal such as RGB of pixels in a continuous area as a center. And generate noise. It is desirable that the value of the standard deviation in the case of the normal distribution function be appropriately determined according to the target image.For example, in the case of a 256-level image, a value of about 100 to 180 For an image, it is 3 to 6 when the image is a face such as a human face or an object, and is 7 to 10 when a natural image such as a landscape generally has no surface. If it is a uniform distribution function, it is not necessary to set the standard deviation, but it is necessary to set the noise level.

Although the average value of each color signal such as RGB of pixels in the continuous area is set as the center of the probability function, this may be shifted. For example, shifting to a brighter direction can avoid dull images. It is also possible to shift in the darker direction.

It is desirable that the function to be used, the value related to the width of the distribution, and the center value can be set arbitrarily by the operator by viewing the image. The continuous area is, for example, a predetermined level for each color signal such as RGB. This is an area where pixels in the difference are continuous. The continuous area includes both a large area and a small area, but only the large area needs to be subjected to noise addition processing. Therefore, the additional processing application determination unit determines that noise addition processing is to be performed only on a large continuous area having a number of pixels equal to or greater than a predetermined value in the continuous area.

If it is determined whether noise addition processing should be performed by searching all continuous areas, there is a problem that processing time becomes longer. Therefore, a pixel level histogram creation unit is provided that creates a histogram of the level of each pixel in the image data in parallel with the storage of the image data, and the continuous area detection unit refers to the histogram to determine the frequency of occurrence. It is desirable to search for a continuous area in order from the pixel level with the largest value. In this case, the continuous region detection unit searches for a continuous region only for a predetermined number of pixel levels having a high occurrence frequency or searches for a continuous region only for a pixel level whose occurrence frequency is equal to or higher than a predetermined value. I do.

In the case where the hue differs to some extent at the boundary between adjacent continuous regions, it is possible to remove an unnatural boundary by averaging adjacent pixels. In order to perform the averaging process, an averaging process determining unit that determines whether to perform the averaging process is performed based on image data at a boundary portion of a region adjacent to the continuous region where the noise adding process is performed, and an averaging process is performed. And an averaging processing unit that performs averaging processing on the boundary part determined to be different.The noise addition processing unit generates noise for each image in the continuous area after the averaging processing has been performed. Let me add. BRIEF DESCRIPTION OF THE FIGURES

FIG. 1 is a diagram illustrating an example in which a continuous area having similar colors and levels is determined.

FIG. 2 is a diagram showing the configuration of the image processing apparatus according to the first embodiment of the present invention. You.

FIG. 3 is a flowchart showing a processing routine in the first embodiment.

FIG. 4 is a flowchart showing a noise addition processing routine in the first embodiment.

FIG. 5 is a diagram illustrating a configuration of an image processing apparatus according to a second embodiment of the present invention.

0 ο

FIG. 6 is a diagram showing a configuration of a modification of the second embodiment.

FIG. 7 is a diagram illustrating the averaging process according to the third embodiment of the present invention.

FIG. 8 is a flowchart showing the averaging routine in the third embodiment. BEST MODE FOR CARRYING OUT THE INVENTION

FIG. 2 is a diagram showing the configuration of the image processing apparatus according to the first embodiment of the present invention, FIG. 3 is a flowchart showing the entire noise adding process of the first embodiment, and FIG. This is a flowchart showing noise addition processing. The first embodiment is an example in which low-resolution image data is printed by a high-resolution image forming apparatus (printer). For example, the input image data is 8-bit RGB data. 6 gradations can be expressed. Each part in Fig. 2 is formed as a processing unit in the computer.

First, in step 101, the user selects whether or not to perform noise addition processing on data. If not in the noise addition processing mode, the image data is sent directly to the binary / multi-value processing section 18 as normal processing, and the result is stored in the print data memory 19 and printed. If the noise addition processing mode has been selected, in step 102, the image data is stored in a bitmap format image data memory 1 for processing. Expands to 1. At the same time, in step 103, the pixel level histogram creation unit 1 checks the level (RGB level) of each pixel of the input image data and counts the number of pixels at each level. To create a histogram. In the present embodiment, it is determined whether or not there is a continuous area in descending order of the number of pixels at a level where the number of pixels is a predetermined number or more from the histogram.

In step 104, the noise addition level is determined. This determines how much noise is added depending on the target image. If a normal distribution function is used, the standard deviation is determined, and if it is a uniform distribution function, the width is determined. For example, the standard deviation is about 3 to 6 for a face such as a human face or an object, and about 7 to 10 for a natural image such as a landscape where there is generally no face. .

In step 105, the level of the maximum number of pixels is selected from the uniform region extraction processing unit 13 power histogram, and in step 106, it is determined whether the number of pixels is equal to or greater than a predetermined value. In this case, a width may be set for a certain level, and the number of pixels within the width may be targeted. If the number of pixels is not equal to or larger than the predetermined value, the process ends. This is because it is determined that there is no large continuous area. Such a determination condition is stored in the processing execution determination condition information memory 15 in advance, and is provided to the uniform region extraction processing unit 13 via the region information memory 14.

If the number of pixels is equal to or larger than a predetermined value, a continuous area is searched in step 107. This is a continuous area of selected levels, ie, the same RGB data or a series of similar data with only one or two levels of RGB values (R ± l, G ± l, B 土 1, etc.). To explore. There may be more than one such continuous region for a level. In step 108, it is determined whether such a continuous area is equal to or larger than a predetermined size. This is because the number of pixels in each continuous area is greater than This is performed by determining whether Since the noise addition processing is more effective when applied to a continuous area having a certain size or more, even a continuous area, a small area is excluded from the target. Information on the continuous area determined to be subjected to the noise addition processing is stored in the clipped image memory 16.

In step 109, the noise generation processing unit 17 performs a process of adding noise in each continuous region determined to perform the noise addition process. The part that does not perform noise addition processing is sent to the binary / Z-multivalue processing section 18 as it is. There are various methods for this processing. Here, an example of generating noise using a normal distribution function will be described with reference to FIG.

In step 121, the average of the image levels of the pixels in the target continuous area is calculated for each of the RGB data. In step 122, the standard deviation of the normal distribution determined in step 104 is set. Steps

1 2 3 sets the first pixel, generates random numbers for each of R, G, and B according to the normal distribution function in step 1 24, and adds random numbers to each RGB value in step 1 25 The data is processed by the Z multi-value processing section 18 and stored in the print data memory 19. This is performed for all pixels in the continuous area. For example, if the image data (RGB) in the continuous area is

When (1 2 8, 1 2 8, 1 2 8), the random numbers generated according to the normal distribution function are (1-5, _ 6, 1), (2, — 5, 1), (3,-2 , 3),…, the corrected data is (1 2 3, 1 2 2, 1 2 7) ヽ (1 3 0, 1 2 3, 1 2 9), (1 3 1, 1 2 6, 1 2 5). The above is the noise addition processing. The noise given by the noise adding process can be modified in various ways depending on how the image is changed, and those parameters can be stored in advance in the processing execution determination condition information memory 15. The power to remember The processing in the noise generation processing unit 17 may be changed accordingly. For example, it is possible to select processing such as shifting the average value (the average value of the standard deviation) in the brighter direction with respect to the average value of the levels in the continuous area. Conversely, it is also possible to shift in a long direction. It is desirable that this shift amount can be set interactively by the user while watching the screen.

In step 110, the level having the next largest number of pixels after the histogram is selected, and steps 106 and thereafter are repeated. In this way, it is determined whether there is a continuous area to be processed at a level having a relatively large number of pixels in the histogram, and noise addition processing is performed on the continuous area to be processed. The stored image data is stored. Thereafter, printing is performed according to the image data.

FIG. 5 is a diagram illustrating a configuration of a video image processing device according to a second embodiment of the present invention. Although the first embodiment is an apparatus for processing image data to be printed, the second embodiment is an apparatus for processing video data to be displayed on a display device.

The video data is temporarily expanded and stored in the bitmap format temporary storage memory 21. For the image data held in the temporary storage memory 21, the conditions stored in the processing execution determination condition information memory 23 by the uniform area cutout processing unit 22 are the same as in the first embodiment. Then, a continuous area to be processed is searched for, and data indicating the continuous area to be processed is stored in the continuous area data memory 24. In the continuous area data memory 24, the average value of the image level of each continuous area necessary for the noise addition processing is also stored.

As in the first embodiment, the noise generation processing unit 25 corrects each pixel in the continuous area based on the data stored in the continuous area data memory 24, and displays the corrected data in the display data memory. Store in memory 26. here The data stored in the display data memory 26 can be switched between the data after correction and the data stored in the uncorrected temporary storage memory 21. It is possible to switch between data that has been processed and data that has not been processed.

As shown in FIG. 6, instead of the display data memory 26, a difference holding memory 27 holds only data of a continuous area processed by the noise generation processing unit 25 as a difference. Then, a combining circuit 28 that combines the image data that has not been processed with the data held in the difference holding memory 27 may be provided. In this case, the memory capacity can be reduced.

In the case where the hue is somewhat different at the boundary between adjacent regions, an averaging process is performed between pixels adjacent at the boundary to remove an unnatural boundary line. Generally, the averaging process is performed by averaging data of adjacent pixels for each of RGB. The averaging process and the noise addition process of the present invention can be performed together. In this case, after performing the averaging process at the boundary of the continuous region, the noise adding process is performed within the continuous region.

FIG. 7 is a diagram for explaining an example of processing in a case where noise addition processing is performed after averaging processing is performed. Here, it is assumed that the averaging process is performed on a portion that is separated by 10 or more at any level of RGB from a region where the pixel value of the pixel is adjacent to the adjacent region.

In Fig. 7, the area shown by the broken line is the continuous area where the noise addition processing is performed, and the RGB values of the pixels in this area are all (124, 152, 189). The RGB values of the pixels in the left area are all (1 2 7, 1 5 5, 1 9 2), and the RGB values of the pixels in the right area are all (1 2 0, 1 5 8, 1 9 4) ), And neither averaging process is performed. Note that the force (not shown) All of the raw RGB values are (127, 155, 192), and no averaging is performed.

The RGB values of the pixels adjacent to this area on the upper side are (13 2, 1 45, 1 19), (1 3 1, 1 4 4, 1 18), (1 2 7, 1 4 1, 1 1 2), (1 2 1, 1 3 4, 1 6), (1 1 6, 1 2 8, 1 16), (1 2 4, 1 3 6, 1 2 4), (144, 154, 158) and (162, 173, 177), and the averaging process is performed. There are various averaging methods, but in this case, we simply calculate the average value of the two adjacent pixels above and below, and use that value as the value of the two pixels. That is, two vertically adjacent pixels have the same value. As a result, the RGB values of the upper pixel are, in order from the left side, (128, 149, 154), (128, 148, 154), (126, 146) , 1 5 1) ヽ (1 2 3, 1 4 3, 1 4 8), (1 2 0, 1 4 0, 1 5 3), (1 2 4, 1 4 4, 1 5 7), (1 3 4, 15 3, 17 4), (14 3, 16 3, 18 3). After performing such averaging, noise addition processing is performed on each pixel within the area indicated by the broken line. For example, the result of adding noise to the above averaged value in accordance with the normal distribution with a standard deviation of 4 is, for example, (1 19, 144, 1 47), (1 1 8, 155, 155), (127, 149, 156), (119, 140, 144), (129, 141, 144) 5), (127, 141, 154), (136, 154, 169), and (153, 163, 184).

FIG. 8 is a flowchart showing the processing in the case where the noise addition processing is performed after the averaging processing is performed.

In the first embodiment and the second embodiment, processing is performed until a continuous region to be processed is determined. Then, in step 131, the area around the continuous area is examined, and in step 132, it is determined whether there is a boundary that requires averaging. You. If the averaging process is required, the averaging process is performed in step 133, and then the noise adding process similar to that described in the first and second embodiments is performed in step 134. If the averaging process is not required, the process proceeds to step 134 to perform the noise adding process.

Although the embodiments of the present invention have been described above, it is needless to say that various modifications can be made to the present invention. Industrial applicability

ADVANTAGE OF THE INVENTION By this invention, it is possible to improve the resolution at the time of reproduction | regeneration, to perform enlargement processing, and to reduce the unnatural impression by the same color and the continuation of the level which arises when reproducing compressed image data. , And an image that gives a more natural feeling can be generated.

The present invention can be applied to multi-valued image data to be printed by a printing device such as a printer or multi-valued image data to be displayed on a display, and image quality can be improved.

Claims

The scope of the claims

1. Detect a continuous area having a level difference of not more than a predetermined value in an image, and determine whether to perform noise addition processing on the detected continuous area, and

An image processing method, characterized in that noise is generated and added to each image in the continuous area where noise addition processing is performed.

2. An image data memory that holds the image data,

A continuous region detecting unit that detects a continuous continuous region having a level difference equal to or less than a predetermined value in the image from the image data;

An additional processing application determining unit that determines whether to perform a noise adding process on the detected continuous region;

An image processing apparatus, comprising: a noise addition processing unit that generates noise and adds noise to each image in the continuous area that performs noise addition processing.

3. The image processing apparatus according to claim 2, wherein

The image processing device, wherein the additional processing application determination unit determines that the noise addition processing is to be performed on a large continuous area having a number of pixels of a predetermined value or more in the continuous area.

4. The image processing apparatus according to claim 2, wherein

A pixel level histogram creating unit for creating a histogram of the level of each pixel in the image data;

The image processing device, wherein the continuous region detection unit searches for a continuous region in descending order of pixel frequency with reference to the histogram.

5. The image processing apparatus according to claim 4, wherein

The continuous area detection unit is configured to detect a predetermined number of pixel levels having a high frequency of occurrence. An image processing apparatus that searches for a continuous area for.

6. The image processing apparatus according to claim 4, wherein

The image processing device, wherein the continuous area detection unit searches for a continuous area for a pixel level whose occurrence frequency is equal to or higher than a predetermined value.

7. The image processing apparatus according to claim 2, wherein

An averaging processing determination unit that determines whether to perform averaging processing from image data of a boundary part of an area adjacent to the continuous area where noise addition processing is performed, and an averaging processing unit that determines that averaging processing is performed. Averaging unit that performs the processing,

The image processing device, wherein the noise addition processing unit generates and adds noise to each image in the continuous area after the averaging process is performed.