JP2010081350A - Method and program for generating oil painting tone image - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide techniques in which even a user not familiar with image processing can obtain an oil painting tone image without any unnatural finish irrespective of the content of an original image. <P>SOLUTION: A method for generating an oil painting tone image includes: a step of: acquiring an original image as an input; an undercoating step of separating the input into a density component and a color component, reducing the number of gradations of the density component to a predetermined number of gradations allowing each gradation level to be visually recognized, and generating an undercoat image composed of the density component having the reduced number of gradations and the color component; a drawing step of emphasizing an outline of the input, smoothing a region excluding the outline portion, and generating a stroke image; and a superposition step of superposing the generated undercoat image and the stroke image, each of the step being performed by a computer. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a generation method and a generation program for generating an oil-tone image by performing image processing on an original image.

In recent years, with the spread of personal computers and digital cameras and higher performance, opportunities to come into contact with digital images are increasing. At the same time, digital image processing and editing using image editing software such as Photoshop (software from Adobe Systems) and Painter (software from COREL CORPORATION) is spreading not only to professionals but also to general users. . Using these image editing software, high-level computer graphics can be created.
However, conversion and drawing of digital images using them requires specialized knowledge and techniques, and sometimes requires a lot of labor.
Therefore, it may be more useful for users who are unfamiliar with the handling of image editing software to easily obtain results using an automated image conversion filter, even if their functions are limited. unknown.

  Several processes have been proposed for the process of converting live-action photographs and videos into painting style (non-photo-realistic rendering or NPR) (see, for example, Patent Documents 1 and 2). Non-photorealistic rendering is routinely used for movies and advertising.

In Patent Document 1, a filter layer formed by overlapping a plurality of white lines inclined at 45 ° and an additional decorative layer on which predetermined characters or the like are superimposed are superimposed on photo data captured by a scanner, and a pencil drawing-like tone is obtained. A method for generating an image or the like is disclosed.
Patent Document 2 discloses a technique for generating a painting-like image by synthesizing an original image with a filter image created with a predetermined shading pattern.
Japanese Patent Laid-Open No. 10-243211 JP 2002-298136 A

As described above, in order to make it easy for a user unfamiliar with the handling of image editing software, the contents of images such as people and landscapes are not limited, that is, they are highly versatile and the process of processing is simple. Therefore, high-speed processing is required. In addition, because of the characteristics of non-photorealistic rendering, what is “visually interesting” is required.
By applying the conventional methods such as Patent Documents 1 and 2, an intended result may be obtained for a certain application. However, since non-photo-realistic rendering has creative aspects, the above-mentioned “visually interesting” value varies from case to case depending on applications and user preferences. From various viewpoints, a new and useful non-photorealistic rendering technique is demanded from both a technical viewpoint and an entertainment viewpoint.

  The present invention has been made in consideration of the above-mentioned circumstances, and observes the impression that the texture and the unique touch of the oil painting are well expressed, in other words, the impression that it was mechanically processed. The goal is to generate an oil painting-like image that is not given to the user and that does not impair the characteristics of the motif of the original image. In addition, the present invention provides a technique that enables even an unfamiliar user of image processing to obtain an oil painting-like image without unnaturalness, regardless of the content of the original image.

  According to the present invention, an original image is acquired as an input, the input is separated into a density component and a color component, and the density component is reduced to a predetermined number of gradations such that each gradation level can be visually recognized. An undercoating step for generating an undercoating image composed of a reduced density component and the color component; a drawing step for generating a stroke image by enhancing the contour of the input and smoothing an area excluding the contour portion; There is provided a method for generating an oil painting tone image, comprising: a superimposing step of superimposing a generated undercoat image and a stroke image, and a computer executing each step.

  Further, from a different point of view, the present invention separates the input image processing into a density component and a color component, and obtains the density component into a predetermined number of gradations such that each gradation level can be visually recognized. A stroke image is generated by generating an undercoat image that includes a subtracted density component with a reduced number of gradations and the color component, and smoothing a region excluding the contour portion while enhancing the input contour. An oil painting tone image generation program is provided that causes a computer to execute a drawing process to perform and a superimposition process to superimpose a generated undercoat image and a stroke image.

  Since the oil painting tone image generation method of the present invention includes an undercoating process for generating an undercoating image and a drawing process for generating a stroke image, each process of the undercoating process and the drawing process corresponding to an oil painting drawing procedure By superimposing the undercoat image and the stroke image created in step 1, an oil painting-like image free from unnaturalness can be obtained regardless of the content of the original image. In addition, it is possible to obtain an image in which the texture and brushstroke unique to oil paintings are well expressed.

  In the present invention, the density component is obtained by removing hue information from a color image. In the field of image processing, density components can be obtained as a result of so-called gray scale conversion. Further, information corresponding to the density component can be obtained as the Y component as a result of YUV conversion. Furthermore, information corresponding to the density component can be obtained as the L component as a result of the Lab conversion.

  On the other hand, the color component is information relating to the hue of the color image. For example, the U and V components as a result of the aforementioned YUV conversion provide color information. Alternatively, the a and b components as a result of the Lab conversion provide color information.

  The process of reducing the number of gradations is for reproducing the texture of rough undercoat in the oil painting production process. In the present invention, since the number of gradations is reduced from the grayscale image, an unnatural hue (false color) that does not exist in the original image compared to the case where the number of gradations of each RGB color component is reduced. Occurrence can be suppressed.

The stroke image is intended to reproduce the stroke of an oil painting. For this purpose, the outline of the original image is retained and emphasized, while the process of reducing the shade to some extent is performed.
Thus, one characteristic aspect of the present invention is that a plurality of images are generated in correspondence with the pictures drawn at each stage of the oil painting creation process, and these images are superimposed.
Hereinafter, preferred embodiments of the present invention will be described.

  In the drawing step, a Kuwahara filter may be applied to the input, and contour enhancement processing may be performed. As described above, the drawing process is a process for generating a stroke image resembling the stroke of an oil painting. According to this aspect, by applying the Kuwahara-filter to the input, it is possible to smooth the input while maintaining the contour, and further, by emphasizing the contour of the image after smoothing, the stroke of the oil painting can be improved. Can be reproduced.

  Furthermore, the drawing step may extract a contour from the image, and further superimpose the extracted contour on an image to which the quahara filter and the contour enhancement processing are applied. In this way, by further superimposing the contour image on the image to which the Kuwahara filter and the contour enhancement processing are applied, it is possible to obtain a stroke image more resembling an oil painting.

  In the undercoating step, the input density component is smoothed, the smoothed density component is reduced to a predetermined number of gradations to obtain a monochrome posterized image, and the input is converted into a YUV color system. A color component composed of U and V components may be obtained, and an undercoat image composed of the density component of the monochrome posterized image and the color component may be generated. The undercoating process is a process for generating an undercoating image drawn with rough shading and outlines. According to this aspect, an undercoat image can be reproduced by applying the color component to the monochrome posterization image. That is, by extracting the density component of the input and reducing the number of gradations thereof, a rough monochrome posterization image is obtained, and the color components U and Y obtained by YUV conversion of the input are obtained. By providing a monochrome posterized image, an undercoat image can be reproduced.

Furthermore, the result of gray scale conversion of the input may be the density component.
Alternatively, a Y component obtained by converting the input into a YUV color system may be used as the density component of the input. The Y component gives an input luminance component, in other words, a density component. Therefore, by performing YUV conversion on the input, it is possible to obtain density components and color components necessary for generating an undercoat image at a time.

  Further, in the superimposing step, a ground pattern of an oil painting canvas prepared in advance may be superimposed in addition to the undercoat image and the stroke image. In this way, the ground pattern of the canvas is reflected in the finished image, and an image more similar to the characteristics of the oil painting is generated.

Furthermore, the superimposing step may perform gradation correction by applying a predetermined gradation characteristic to the superimposed image. By correcting the gradation characteristics of the superimposed image, it is possible to obtain an oil painting-like image having no unnaturalness with respect to the original image.
The various preferable aspects shown here can also be combined.

Hereinafter, the present invention will be described in more detail with reference to the drawings. In addition, the following description is an illustration in all the points, Comprising: It should not be interpreted as limiting this invention.
The present invention generates a plurality of images corresponding to the pictures drawn at each stage of the oil painting creation process, and superimposes these images. Therefore, prior to the description of this embodiment, the characteristics of the oil painting and its drawing procedure will be briefly described.

≪Basic knowledge of oil painting≫
1.1 Features of oil painting Oil painting is drawn with oil paints obtained by kneading oil and pigment, which are colorants. Oil paints are transparent due to the difference in refractive index between the oil and the pigment and the particle size of the pigment itself. Therefore, in oil paintings, it is possible to paint transparently reflecting the color of the groundwork. The pigment floats in the oil as the color extender, and light passes through the gap between the pigment. Such a transparent structure is called a multi-layer coloring scheme, and is one of the major features of the technique of oil painting. An example of a multi-layered coloring scheme is Johannes Vermeer (1632-1675) “Woman and two men”. In this work, the effect of making the red of the dress brighter is obtained by using Ultra Marine Blue as the base of the red dress.

1.2 Oil Painting Drawing Procedure The oil painting drawing procedure is roughly divided into the following four steps.
[step1] Preparation of the canvas The part that physically holds the foundation and the paint layer is called the support. In oil paintings, this is often the canvas. Pre-processing such as erasing the color of the support itself or drawing a drawing is performed as necessary (FIG. 1A).
[step2] Undercoat Draw the light and darkness of the motif with a four-step gradation using dark brown paint (Burnschenna) along the drawing. From the top, a unique motif color is applied like a watercolor (Fig. 1 (b)).
[step3] Draw Reproduce the colors, lightness and darkness of the motif using a larger amount of paint than the undercoat stage. By using a large amount of paint, a unique painting with a brush is born (Fig. 1 (c)).
[step4] Finish Perform a partial rework. In order to prevent discoloration and fading, a storage varnish may be applied (FIG. 1 (d)).

≪Each process of image processing≫
Next, a procedure for reproducing the above-described oil painting drawing procedure on a computer in a pseudo manner will be described. Specifically, the procedure is to create a canvas image, an undercoat image, and a stroke (after brush painting) image, and then synthesize them. In addition, the procedure described in this embodiment is realized by the inventor based on the above-described multi-layer coloring method.
The specifications of the computer used for image processing in the following embodiment are as follows: OS is Microsoft Windows XP (registered trademark), CPU is Intel Pentium (registered trademark) 4, operating frequency is 2.53 GHz, memory capacity is It's 512MB.

2.1 Outline of this embodiment In this embodiment, an original image is processed in the following procedure to obtain an output.
[step1] Preparation of the canvas Capture the canvas pattern image data.
[step2] Create an undercoat image from the original image of the undercoat process.
[step3] Create a stroke image from the original drawing process image.
[step 4] Finish The images related to step 1, step 2, and step 3 are superimposed and an appropriate correction is performed to obtain an output image.

2.2 Original Image Three types of “person” (see FIG. 2), “animal” (see FIG. 3), and “scenery” (see FIG. 4) shown in FIGS. Both elements are likely to be subjects of photographs, and the versatility of the method of the present invention is confirmed by comparing each element.
In this embodiment, the variable block is used below. The value of block affects the application area of various filters. The larger the value, the wider the application area of the filter. The value of block is determined by the following formula (1) according to the number of pixels of the original image.

(1) 2.3 [step1] Preparation of canvas A commercially available canvas was scanned to obtain a canvas image. It is a bitmap image of 2400 × 2400 pixels (see FIG. 5).

2.4 [step2] Undercoat process Create an undercoat image from the original image using the following procedure.
1. Gaussian filter
2. Grayscale conversion
3. Posterization
4. Composition with original image using YUV color system The details of each procedure are described below.

2.4.1 Gaussian filter The Gaussian filter is a weighted smoothing filter using a normal distribution. The normal distribution G (i) of the pixels i apart from the target pixel is given by equation (2), where σ is the standard deviation.

(2) Further, when the size of the local region centered on the target pixel is expressed as (2N + 1) × (2N + 1), it is assumed that N = 3σ. This is because the normal distribution almost converges at 3 times σ.

First, a weighted average in the horizontal direction is performed. If block = 1, the area corresponding to FIG. 6 is targeted.
Here, the pixel P3 indicated by reference numeral 11 is the target pixel. The pixel value P3 ′ after the weighted average is obtained by equation (3).

(3) Subsequently, the weighted average in the vertical direction is performed in the same procedure.

  The smoothed image shown in FIGS. 7-9 was obtained by the above process. The weighted average based on the two-dimensional normal distribution according to the distance is equivalent to applying the weighted average based on the one-dimensional normal distribution in the horizontal and vertical directions, and is mathematically proved by F. Waltz and W. Miller. (See Frederick M. Waltz, John WV Miller, “An efficient algorithm for Gaussian blur using finite-state machines,” Proc. SPIE Conf. On Machine Vision Systems for Inspection and Metrology VII pp. 3521-37 (1998)).

2.4.2 Grayscale conversion Convert the smoothed image obtained in the previous item 2.4.1 to a grayscale image. If the R, G, and B values in each pixel of the smoothed image are converted to R ′, G ′, and B ′, respectively, Equation (4) is obtained.

(4) From the above, the gray scale images of FIGS.

2.4.3 Posterization of a grayscale image Posterization is a process of arbitrarily changing the number of gradations in an image. In this research, in order to obtain a smoother contour line, the average value of the pixel values in the local region is taken as preprocessing. The size of the local area is (2block + 1) × (2block + 1).
Subsequently, the average value P of the obtained pixel values is converted into a pixel value P ′ in (5).

(5) As a result of processing with the number of gradations pos = 8, monochrome posterized images, FIGS.

2.4.4 Composition with original image using conversion to YUV color system The YUV color system has a luminance signal Y, a luminance signal and blue difference signal U (Cb), and a luminance signal and red difference signal V (Cr ). Equation (6) shows the conversion formula from the RGB color system.

(6)

The luminance signal (Y) is obtained from the posterization image in item 2.4.3, and the color information (U, V) is obtained from the original image. However, since the R, G, and B values are the same in the grayscale image, now Y = R = G = B. By converting these YUV values into R, G, and B values, the undercoat images having the outline of the posterization image and the color of the original image in items 2.4.3 were obtained, as shown in FIGS.
In this embodiment, a gray scale image is generated from an RGB color system original image and processed to obtain a monochrome posterized image. However, the luminance information Y of the YUV color system is used to generate a monochrome / colorized image. A posterization image may be generated.
Expression (7) shows a conversion formula from the YUV color system to the RGB color system.

(7)

2.5 [step3] Drawing process Next, the detailed procedure of the drawing process is explained. In the drawing process, a stroke image is created from the original image in the following procedure.
1. Kuwahara filter
2. Unsharp mask
3. Contour filter
4. Multiply synthesis The details of each procedure are explained.

2.5.1 Kuwahara filter
The Kuwahara filter is a type of Maximum Homogeneity Neighbor (MHN) filter and is characterized by smoothing while maintaining the contour (for example, "MemoNyanDum: C # GDI +", [online], [June 18, 2006 Day search], Internet <URL: http://junki.lix.jp/csgdip.html>).

The Kwahara filter first provides four (block + 1) × (block + 1) local regions centered on the pixel of interest. If block = 2, the result is as shown in FIG. The central pixel indicated by reference numeral 21 is the target pixel, and each of the four local regions 23, 24, 25, and 26 is arranged with one vertex as the target pixel.
Subsequently, the maximum value and the minimum value of each value of R, G, and B in each of these four regions are obtained, and the differences are summed. The region where the total value of these four is the smallest has the maximum uniformity in these four regions. Average values of R, G, and B values in the maximum uniformity region are obtained, and the obtained average value is adopted as the value of the target pixel. The image of FIGS. 20-22 was obtained by applying this Kuwahara filter.

2.5.2 Unsharp mask
The unsharp mask is a process that enhances the outline and contrast and sharpens the image. The pixel value difference between the target image and the smoothed image is added to the target image.
In this embodiment, the difference between the output image of item 2.5.1 and the smoothed image of the original image obtained in 2.4.1 is used. That is, if the pixel values in these corresponding pixels are P0 and P1, respectively, the pixel value P2 of the output image can be obtained by equation (8).

(8) As a result of processing at per = 150, FIGS.

2.5.3 Extraction of contour line by contour filter The contour filter is a filter that extracts the contour by measuring the distance of the color, considering each value of R, G, B as primary independent orthogonal coordinates ( (See FIG. 26). In Figure 26, the P ₁₀ indicated by a symbol 31 is a pixel of interest.
First, the color distances D between the pixel of interest and its upper right, right, lower right, and lower pixels are determined by Equation (9).

(9) Subsequently, the maximum value max of D ₀₁ , D ₁₁ , D ₂₁ , D ₂₀ is obtained, and the R, G, B values of the target pixel are determined as follows.

(10) From the above, FIGS.

2.5.4 Multiply synthesis
Synthesize the output image of 2.5.2 and the contour image obtained in 2.5.3. If the pixel values in these corresponding pixels are P0 and P1, respectively, the pixel value P2 of the output image can be obtained by equation (11).

(11) From the above, stroke images, FIGS.

2.6 [step4] Finishing In the finishing step, an output image is obtained by the following procedure.
1. Superimposition of output image at [step1], [step2], [step3]

2. Gamma correction
2.6.1 Superimposition of output images in [step1], [step2], [step3] The canvas image obtained in item 2.3, the undercoat image obtained in item 2.4.4, and the stroke image obtained in item 2.5.4 A superimposed image is obtained. When the pixel values in these corresponding pixels are P0, P1, and P2, respectively, the pixel value P3 of the superimposed image is obtained by Expression (12).

(12) From the above, superimposed images, FIGS.

2.6.2 Gamma correction Gamma correction is widely used for brightness correction between image input / output devices.
Since the integrated image obtained in the previous item 2.6.1 is darker than the original image because it was synthesized with the canvas image, it is corrected. If the R, G, and B values are converted into R ′, G ′, and B ′, respectively, Equation (13) is obtained.

(13)
As a result of processing with γ = 2.0, an oil painting-like output image, FIGS.

≪Processing procedure≫
Here, the flow of each process for generating the above-mentioned oil painting tone image will be described again.
FIG. 43 is an explanatory diagram showing the flow of each step of image processing according to this embodiment. This will be described with reference to FIG.
First, an original image as input is read (step S11). Sample images are shown in FIGS. The read image is used to generate an undercoat image and a stroke image.

In the step of generating an undercoat image (undercoat step), the input is smoothed using a Gaussian filter to obtain a smoothed image (step S21). Sample images are shown in FIGS.
Subsequently, the smoothed image is subjected to gray scale conversion (step S23). Sample images are shown in FIGS.
Then, the number of gradations of the grayscale-converted image is reduced to a predetermined number (step S25). Sample images are shown in FIGS.
On the other hand, the input is YUV converted to obtain U and V components as color components (step S27).

The color component of step S27 is applied to each region of the posterization image of step S25 to obtain an undercoat image (step S29). Sample images are shown in FIGS.
As a modification, the luminance component (Y component) obtained by the YUV conversion in step S27 may be used as the input in step S21. In this case, the gray scale conversion of S23 is not necessary.

On the other hand, in the drawing process for generating the stroke image, the input is smoothed by using the Kuwahara filter (step S41). Sample images are shown in FIGS. The image smoothed by the Kuwahara filter is characterized in that the contour line is retained compared to the smoothed image by the Gaussian filter.
Then, the contour line is further emphasized using an unsharp mask on the image after the application of the Kuwahara filter (step S43). Here, the unsharp mask uses the difference between the smoothed image obtained by the Kuwahara filter in step S41 and the smoothed image obtained by the Gaussian filter in step S21. Sample images are shown in FIGS.
Further, a contour filter is applied to the input to extract a contour (step S45). Sample images are shown in FIGS.

The obtained contour is combined with the image obtained by the unsharp mask in step S43 to obtain a stroke image with a clearer contour (step S47). Sample images are shown in FIGS.
As a modification, steps S45 and 47 may be omitted, and the unsharp mask result may be a stroke image. However, it is preferable to perform the processing of steps S45 and 47 in order to reproduce a brush-like touch that is more oily.

The undercoat image and stroke image obtained as described above are superimposed. Alternatively, a campus pattern image prepared in advance may be read (step S13) and superimposed with the undercoat image and the stroke image (step S61). Sample images are shown in FIGS.
Then, gamma correction is performed to adjust the gradation characteristics of the image obtained in step S61 (step S63). The gamma correction may be performed using a predetermined characteristic. Alternatively, a histogram of the image obtained in step S61 may be taken and characteristics may be determined based on the result.
The image obtained as described above is output as a processing result. Sample images are shown in FIGS.

≪Comparison with processing result of commercial software≫
In this chapter, a comparison is made between a commercially available software product and this embodiment. Typical image editing software includes Adobe Systems' Photoshop and Corel's Painter. In this study, the former is used for comparison. This is because automatic conversion to painterly by Painter has strong individuality such as Van Gogh touch and Surah touch, and in comparison, Photoshop has many conversion filters that can be combined for general purposes.

3.1 Procedure
Using Photoshop CS2 from Adobe Systems, filter processing that is relatively close to each stage of this embodiment is performed in order. The image shown in FIG. 2 is used as the original image.
1. Texture riser
2. Posterization
3. Palette knife, unsharp mask
4. Procedure 1-3 Integration of layers and brightness correction Output images at each stage are shown in FIGS. 39 shows an image after applying the textureizer, FIG. 40 shows an image after applying posterization, and FIG. 41 shows an image after applying the palette knife and unsharp mask. FIG. 42 shows an image obtained by superimposing the images of the layers 1 to 3 and correcting the brightness.

3.2 Study Compared with this embodiment, PhotoshopCS2 filter conversion has several problems.
First, since the textureizer (FIG. 39) has regularity in the texture arrangement, it has a mechanical impression. Therefore, it is not suitable for oil painting tone conversion. In posterization (FIG. 40), unnatural color irregularities are sometimes found. This is probably because the number of gradations is individually converted for each value of R, G, and B. In addition, the color boundaries are not smooth. Attempting posterization with Corel's Painter Essentials 3, the result was similar to PhotoshopCS2.
Subsequently, the pallet knife (FIG. 41) has more severe damage to the contour line compared to the Kuwahara filter.
Regarding the integration of layers and correction of brightness (FIG. 42), the ratio of manual work is large and time-consuming.

  Photoshop is a versatile image processing software and is not specialized for filtering functions. In addition, the present invention is not specialized in oil painting-like image generation as in the present invention. Therefore, it is natural that manual work is increased as compared with this embodiment for automatic conversion. However, each stage of this embodiment, the resulting image appears to be comparable to the commercial product.

  As described above, it has been confirmed that the method of the present invention can convert a photograph into an oil painting-like image without any auxiliary input from the user, and has versatility regardless of the type of photograph. Moreover, in terms of automatic conversion to an oil painting-like image, it has higher performance than a commercially available product.

  Future challenges include unnatural fragmentation resulting from the use of thresholds for posterization (see, for example, the sky and sea portions of FIG. 15). If the number of gradations after posterization is set to a high value such as 32, this problem can be generally avoided. If priority is given to a more natural finish, it is preferable to determine the optimum gradation number of the target image after actually outputting the result image. However, this makes the user's operation complicated. For this reason, improvements such as fixing the number of gradations after posterization to 32 gradations that are more stages than in the embodiment, or automatically discriminating the optimum number of gradations by adding some processing can be considered.

  In addition to the embodiments described above, there can be various modifications of the present invention. These modifications should not be construed as not belonging to the scope of the present invention. The present invention should include the meaning equivalent to the scope of the claims and all modifications within the scope.

It is explanatory drawing which shows the example of the drawing procedure of the oil painting based on the idea of this invention. It is explanatory drawing which shows an example of the input image which concerns on this embodiment (a person, 1500x2000 pixel: block = 5). It is explanatory drawing which shows a different example of the input image which concerns on this embodiment (animal, 2000x1500 pixel: block = 5). It is explanatory drawing which shows another example of the input image which concerns on this embodiment (landscape, 2000x1500 pixel: block = 5). It is explanatory drawing which shows an example of the canvas image which concerns on this embodiment. It is explanatory drawing which shows the application pixel of the horizontal direction of the Gaussian filter which concerns on this embodiment. It is explanatory drawing which shows an example of the smoothed image which concerns on this embodiment (person). It is explanatory drawing which shows an example from which the smoothed image which concerns on this embodiment differs (animal). It is explanatory drawing which shows another example of the smoothed image which concerns on this embodiment (landscape). It is explanatory drawing which shows an example of the gray scale image which concerns on this embodiment (person). It is explanatory drawing which shows an example from which the gray scale image which concerns on this embodiment differs (animal). It is explanatory drawing which shows a further different example of the gray scale image which concerns on this embodiment (landscape). It is explanatory drawing which shows an example of the monochrome posterization image which concerns on this embodiment (person). It is explanatory drawing which shows an example from which the monochrome posterization image which concerns on this embodiment differs (animal). It is explanatory drawing which shows another example of the monochrome posterization image which concerns on this embodiment (landscape). It is explanatory drawing which shows an example of the undercoat image which concerns on this embodiment (person). It is explanatory drawing which shows an example from which the undercoat image which concerns on this embodiment differs (animal). It is explanatory drawing which shows another example of the undercoat image which concerns on this embodiment (landscape). It is explanatory drawing which shows the detail of the Kuwahara filter (Kuwahara filter) which concerns on this embodiment. block = 2. The center pixel is the target pixel. It is explanatory drawing (example) which shows an example of the image after the Kuwahara filter which concerns on this embodiment is applied. It is explanatory drawing which shows an example from which the image after the Kuwahara filter which concerns on this embodiment is applied differs (animal). It is explanatory drawing which shows a further different example of the image after the Kuwahara filter which concerns on this embodiment is applied (landscape). It is explanatory drawing which shows an example of the image after unsharp mask application concerning this embodiment (person). It is explanatory drawing which shows an example from which the image after an unsharp mask application concerning this embodiment differs (animal). It is explanatory drawing which shows a further different example of the image after unsharp mask application concerning this embodiment (landscape). It is explanatory drawing which shows the detail of a contour filter. It is explanatory drawing which shows an example of the image after the outline filter application concerning this embodiment (person). It is explanatory drawing which shows an example from which the image after the outline filter application concerning this embodiment differs (animal). It is explanatory drawing which shows a further different example of the image after the outline filter application concerning this embodiment (landscape). It is explanatory drawing which shows an example of the stroke image which concerns on this embodiment (person). It is explanatory drawing which shows an example from which the stroke image which concerns on this embodiment differs (animal). It is explanatory drawing which shows a further different example of the stroke image which concerns on this embodiment (landscape). It is explanatory drawing which shows an example of the superimposed image which concerns on this embodiment (person). It is explanatory drawing which shows a different example of the superimposed image which concerns on this embodiment (animal). It is explanatory drawing which shows another example of the superimposed image which concerns on this embodiment (landscape). It is explanatory drawing which shows an example of the output image which concerns on this embodiment (person). It is explanatory drawing which shows a different example of the output image which concerns on this embodiment (animal). It is explanatory drawing which shows another example of the output image which concerns on this embodiment (landscape). It is explanatory drawing which shows an example of the filter process using commercial software. It is an image after applying the texture riser. It is explanatory drawing which shows an example of the filter process using commercial software. It is the image after posterization application. It is explanatory drawing which shows an example of the filter process using commercial software. It is an image after applying a palette knife and unsharp mask. It is explanatory drawing which shows an example of the filter process using commercial software. It is an image in which brightness adjustment is performed after layer superposition. It is explanatory drawing which shows the flow of each process of the image processing which concerns on this embodiment.

Explanation of symbols

11, 21, 31: pixel of interest 23, 24, 25, 26: local region

Claims (9)

Acquiring an original image as input;
The input is separated into a density component and a color component, the density component is reduced to a predetermined number of gradations such that each gradation level can be visually recognized, and an undercoat image composed of the density component with the reduced number of gradations and the color component An undercoating step to produce
A drawing step of generating a stroke image by emphasizing the input contour and smoothing a region excluding the contour portion;
An oil painting tone image generation method comprising: a superimposition step of superimposing a generated undercoat image and a stroke image, and a computer executing each step.   The generation method according to claim 1, wherein in the drawing step, a Kuwahara filter is applied to the input and an edge enhancement process is performed.   The generation method according to claim 2, wherein the drawing step extracts a contour from the image, and further superimposes the extracted contour on an image to which a Kuwahara filter and a contour enhancement process are applied.   The undercoating step smoothes the input density component, reduces the smoothed density component to a predetermined number of gradations to obtain a monochrome posterized image, and converts the input into a YUV color system to convert U and The generation method according to claim 1, wherein a color component including a V component is obtained, and an undercoat image including the density component of the monochrome posterized image and the color component is generated.   The generation method according to claim 4, wherein the density component is a result of grayscale conversion of the input.   The generation method according to claim 4, wherein a Y component obtained by converting the input into a YUV color system is the density component of the input.   The generation method according to claim 1, wherein the superimposing step superimposes a ground pattern of an oil painting canvas prepared in advance in addition to the undercoat image and the stroke image.   The generation method according to claim 1, wherein the superimposing step performs tone correction by applying a predetermined tone characteristic to the superimposed image. Processing to acquire the original image as input,
The input is separated into a density component and a color component, the density component is reduced to a predetermined number of gradations such that each gradation level can be visually recognized, and an undercoat image composed of the density component with the reduced number of gradations and the color component Undercoat processing to generate
A drawing process for generating a stroke image by emphasizing the input contour and smoothing a region excluding the contour portion;
An oil painting-tone image generation program that causes a computer to execute each of superimposition processing for superimposing a generated undercoat image and a stroke image.