JP2006338365A - Moving image generation device, moving image generation method and moving image generation program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To generate a moving image with a feeling of reality from one image, and to prevent quality deterioration due to numerical error diffusion when using a partial differential equation having advective terms for the generation of a moving image. <P>SOLUTION: A time-series image and velocity field information acquired by using the image are stored in association with each other in a storage device 17, and when the moving image is generated from the image inputted by an input part 11, a time-series image similar to an inputted image is retrieved by a velocity field retrieving part 14, and the velocity field information corresponding to the similar image is substituted in the partial differential equation having time terms or advective terms, and the partial differential equation is processed by using the speed field information suitable for the inputted image, and the moving image with a feeling of reality is generated by a moving image generation part 15. Also, any numerical error is suppressed by solving the partial differential equation by applying a discrete solution by a CIP method. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a technique for generating a time-series image based on a limited video material such as one image.

  In recent years, there has been a rapid increase in technology for generating various new images by giving various expressions to digital images and images. For example, image quality improvement for photographic images is typical. However, there is a technique called morphing for generating an intermediate image from two images as a method for giving motion to an image. For a technique for generating a moving image from one image, the warping method and Patent Document 1 Most are described, and little is known about other methods.

  The warping method performs simple linear interpolation between each frame of an image when trying to give a continuous change to the image. This is because the video producer spends a lot of time on one image and repeats trial and error to make a little movement. However, for general users, it is not possible to take time for trial and error, and it is also necessary for video producers to visually adjust the amount and balance of subtle movements between images. There is a problem that it is low.

  On the other hand, the technique of Patent Document 1 applies a partial differential equation (hereinafter, referred to as “advection equation” as appropriate) having a time term and an advection term using a velocity field vector, and arbitrary time for one image. The images in are sequentially created. The partial differential equation is discretized by the finite difference method and calculated.

While the warping method generates the image of the next frame with a one-to-one correspondence between frames, it remains a mathematical expression, whereas the advection equation corresponds to a plurality of pixels to one pixel between frames. An image is generated by the method, and a physical expression in which information of neighboring pixels is successively reflected for each pixel in the next frame is included. In addition, there are the following as technical documents related to the present application.
JP 2004-334694 A BKPHorn and BGSchunck, "Determining Optical Flow", Artifical Intelligence, vol. 17, no.1, pp.185-203, 1981. Supervised by Hideyuki Tamura, edited by Japan Industrial Technology Center, “Introduction to Computer Image Processing”, Soken Publishing Yoshinori Uesaka, Kazuhiko Ozeki, “Pattern Recognition and Learning Algorithm”, Bunichi General Publishing Tomoharu Nagao, “Optimization Algorithm”, Shosodo HKLee and JHKim, "An HMM-based threshold model approach for gesture recognition", IEEE Trans. PAMI, vol.21, no. 10, pp.961-973, 1999. Toru Nakagawa and Yoshio Koyanagi, “Experimental Data Analysis by Least Squares Method”, University of Tokyo Press, 1995. Keisuke Ikeda, “Nonlinear phenomena in fluids: Mathematical analysis and applications”, Asakura Shoten T. Yabe and T. Aoki, "A universal solver for hyperbolic equations by cubic-polynomial interpolation l. One-dimensional solver", Computer Physics Communications 66, North-Holland, pp. 219-232, 1991.

  However, in Patent Document 1, no consideration is given to giving a realistic motion to an image of an object that moves in a complicated and realistic manner, such as a tree or a water surface.

  In addition, when using the advection equation, repetitive calculation is performed recursively to generate a plurality of images from the next frame on from one image, but the image may be blurred due to the influence of numerical error diffusion, There is a problem that a vibration-like pattern occurs due to the instability reduction. This always occurs especially when speeding up.

  The present invention has been made in view of the above, and an object thereof is to make it possible to generate a moving image with a sense of reality from a single image.

  Another object of the present invention is to prevent image quality degradation due to numerical error diffusion when using the advection equation.

  A moving image generating apparatus according to a first aspect of the present invention includes a storage unit that stores a time-series image and velocity field information obtained using the image in association with each other, an input unit that receives an input of the image, An image similar to the input image is searched from the storage means, a search means for reading out speed field information corresponding to the searched image from the storage means, an input image and the read speed field information, And generating means for generating a moving image by substituting and processing the advection term of the partial differential equation having a time term and an advection term.

  In the present invention, a time-series image and velocity field information obtained using the image are stored in association with each other, and similar to the input image when generating a moving image from the input image. By using the velocity field information that is suitable for the input image, the velocity field information corresponding to the similar image is substituted into the advection term of the partial differential equation having the time term and the advection term and processed. Since the partial differential equation is processed, a moving image suitable for the input image can be generated. The storage means corresponds to, for example, a magnetic storage device, an optical disk device, a memory, or the like.

  Here, the advection term is represented by a product of a velocity component and a spatial first derivative of the image intensity, and inputting the velocity field information to the velocity component ensures the accuracy of moving image generation. Is desirable.

  Further, the search means calculates the similarity between the input image and the time-series image using a correlation function, and searching for a similar image based on the similarity determines the search accuracy of the similar image. Desirable to secure.

  The moving image generation apparatus reads a time-series image from the storage unit, detects speed field information by applying an optical flow method to the image, and associates the speed field information with the time-series image in the storage unit. It has a velocity field detection means for memorizing.

  In the present invention, a moving image with simple motion is obtained by using the velocity field information detected by applying the optical flow method to the time-series image as velocity field information when generating the moving image of the input image. Can be easily generated.

  The moving image generating apparatus reads a time-series image and corresponding velocity field information from the storage means, and substitutes these into an input / output function that outputs velocity field information in response to an image and a coefficient input. Learning, storing the time-series image and the coefficient in association in the storage unit, reading out the coefficient corresponding to the similar image searched by the search unit from the storage unit, and reading out the input image and the input image Learning means for substituting the calculated coefficient into the input / output function to calculate speed field information and store it in the storage means, and the search means reads the speed field information calculated by the learning means from the storage means It is characterized by.

  In the present invention, the time series image and the velocity field information are respectively substituted into an input / output function that outputs velocity field information with respect to the input of the image and the coefficient, and the coefficient is learned. By substituting into the input / output function, more appropriate velocity field information is calculated for the input image, and this velocity field information is substituted into the advection term of the partial differential equation to generate a moving image. Appropriate moving images can be generated for images of various objects.

  In the moving image generating apparatus, the moving image generating means solves the partial differential equation by applying a discrete solution method using a CIP method.

  In the present invention, by solving the partial differential equation by applying the discrete solution method by the CIP method, the value of the adjacent pixel is interpolated by the cubic interpolation function, so that the numerical error can be suppressed.

  In the moving image generating method according to the second aspect of the present invention, a time series image and speed field information obtained using the image are associated with each other and stored in a storage unit, and an input unit inputs an image. A step of receiving, a step of searching the storage unit for an image similar to the input image by the search unit, a step of reading speed field information corresponding to the searched image from the storage unit, and a step of inputting by the generation unit Generating a moving image by substituting and processing the image and the read velocity field information into the advection term of a partial differential equation having a time term and an advection term.

  A moving image generating program according to a third aspect of the present invention includes a process for storing a time-series image and speed field information obtained using the image in association with each other in a storage unit, and an input process for receiving an input of the image A search process for retrieving an image similar to the input image from the storage unit, and retrieving speed field information corresponding to the retrieved image from the storage unit, the input image and the read speed field information And a generation process for generating a moving image by performing processing by substituting into the advection term of the partial differential equation having a time term and an advection term.

  According to the present invention, it is possible to generate a moving image with a sense of reality with respect to one input image. Further, image quality deterioration due to numerical error diffusion can be prevented.

  FIG. 1 is a block diagram illustrating a configuration of a moving image generating apparatus according to an embodiment. The moving image generating apparatus 1 shown in FIG. 1 includes an input unit 11, a speed field detecting unit 12, a speed field learning unit 13, a speed field searching unit 14, a moving image generating unit 15, a display unit 16, and a storage device 17. The moving image generating apparatus 1 is configured by a computer including an arithmetic processing device, a storage device, a memory, and the like, and the processing of each unit is executed by a program. This program is stored in the storage device 17 and is read out and executed by the arithmetic processing unit.

  The moving image generating apparatus 1 generates a moving image by processing a partial differential equation having a time term and an advection term in the moving image generating unit 15 for one image input by the input unit 11 as a feature thereof. In this case, as the velocity component in the advection term, learning is performed using velocity field information previously detected by the velocity field detector 12 using another time-series image or another time-series image by the velocity field learning unit 13. By using the speed field information corresponding to the time-series image similar to the one image among the speed field information, a moving image with a sense of reality is generated. Hereinafter, processing in each unit will be described in detail.

  The input unit 11 receives time-series images from a camera or a database and stores them in the storage device 17. Further, the input of one image to be generated as a moving image is received and stored in the storage device 17.

  The speed field detection unit 12 reads a time series image from the storage device 17, applies an optical flow method (see Non-Patent Document 1) to the image to detect speed field information, and converts the speed field information into the time series image. The data are stored in the storage device 17 in association with each other. Specifically, as shown in FIG. 2, the speed field detection unit 12 detects a speed vector indicating the magnitude and direction of the movement of the moving object as speed field information by comparing two images. For example, FIG. 4A shows a wave image at time n, FIG. 4B shows a wave image at time n + 1, and FIG. 4C shows velocity field information obtained using the images of FIGS. Is indicated by an image, and arrows indicate the magnitude and direction of movement at the pixel point. As another example, as shown in FIG. 3, by comparing the image of a small star in FIG. 3A with the image of a large star in FIG. A velocity field vector indicating a change in the shape of is detected. The optical flow method is suitable for detecting velocity field information for a moving object that moves relatively simply.

The speed field search unit 14 searches the storage device 17 for a time-series image similar to this image when the input unit 11 inputs one image to be generated as a moving image, and corresponds to the searched similar image. The speed field information to be read is read from the storage device 17. Specifically, the similarity between the input image and the time-series image is calculated using a correlation function, and an image is searched based on this similarity. For example, an image with the highest similarity is searched. For this correlation function, for example, the normalized correlation function described in Equation (5.34) on page 150 of Non-Patent Document 2 is used. This equation is shown as follows.

In this embodiment, f is an input image, t is a time-series image, and u = v = 0 in the output m ^* (u, v). The output m ^* (0,0) represents the degree of similarity in the range of 0 to 1. When it is 0, it indicates that it is not similar at all, and when it is 1, it indicates that it matches. Usually, a threshold value is set, and for example, when m ^* (0,0)> 0.8, it is determined that they are similar.

  When calculating the similarity using the correlation function, the speed field search unit 14 divides the input image 400 and the time-series image 410 into a plurality of sub-blocks, as shown in FIG. In the figure, as an example, each image is divided into four parts. Then, with respect to one sub-block of the input image 400, the similarity with all the sub-blocks in the time-series image 410 is obtained, and one time-series image with the highest similarity is searched. Subsequently, the speed field information 420 corresponding to the searched similar image is read from the storage device 17 and sent to the moving image generation unit 15.

  The speed field learning unit 13 reads the time series image and the speed field information obtained by applying the optical flow method to the time series image from the storage device 17, and further applies a neural network (see Non-Patent Documents 3 and 4) or the like. Accurate speed field information is obtained and stored in the storage device 17. Specific processing when applying the neural network is as follows.

Here, Formula (8.1) described on page 105 of Non-Patent Document 4 is used. This equation is shown as follows.

Here, n is the number of pixels, and θ is a predetermined threshold value. In the present embodiment, the function f (x) is an input / output function that outputs velocity field information y with respect to the input of the image O and the weight coupling coefficient w. For the input / output function f (x), for example, a sigmoid function of the following equation as shown in Equation (8.7) of Non-Patent Document 4 is used.

  The velocity field learning unit 13 substitutes the time-series image for O in Equation (2), and substitutes velocity field information obtained by the optical flow method for y. As this velocity field information, for example, information that can be expressed as a two-dimensional image having the same number of pixels as the image O as shown in FIG. 5 is used. Then, as described in Non-Patent Document 4, the weight coupling coefficient w is determined by repeating the process until a certain convergence condition is satisfied using, for example, the steepest descent method for Expression (2). This is called the learning process. The weight coupling coefficient w obtained by learning is stored in the storage device 17.

  Then, when one image that is a moving image generation target is input by the input unit 11 and a time series image similar to this image is searched from the storage device 17 by the speed field search unit 14, the speed field learning unit 13 Then, the weighted coupling coefficient w corresponding to the searched similar image is read from the storage device 17, and the input image is substituted for O and the read weighted coupling coefficient is substituted for w for the expression (2). Then, velocity field information y suitable for the input image is calculated and stored in the storage device 17. The speed field search unit 14 reads the speed field information y from the storage device 17 and sends it to the moving image generation unit 15.

  The velocity field learning unit 13 may apply the HMM method (Hidden Markov Model, see Non-Patent Document 5) or the least square method (see Non-Patent Document 6) in addition to the neural network as a coefficient learning method. In the HMM method, elements of the transition matrix correspond to coefficients. In the least square method, a linear sum combination model that outputs velocity field information with respect to a time-series image and a coefficient input is created, and the coefficient is obtained by the least square method. The method of using the coefficient used by each method is the same as the weight coupling coefficient in the neural network described above.

  The moving image generating unit 15 generates a moving image by substituting the input image and the velocity field information read from the storage device 17 into a partial differential equation having a time term and an advection term. The velocity field information used here is obtained by the velocity field detection unit 12 or the velocity field learning unit 13, and is velocity field information corresponding to a time-series image similar to the input image. Whether the velocity field information obtained by the velocity field detector 12 or the velocity field learning unit 13 is used is determined based on the input image. For example, when the motion of the input image is simple, the speed field detecting unit 12 is used, and when the input image is complicated, the speed field learning unit 13 is used.

A partial differential equation having a time term and an advection term is expressed by the following equation (see Non-Patent Document 7).

  Here, the image intensity is defined as I (x, y, n), where (x, y) is the pixel, n is the discretized time. Here, the time coefficient Δt is 1.0. U is a two-dimensional velocity field vector and is given by velocity field information. ▽ is a first-order spatial differential operator, and · is a vector dot product. The first term on the left side of Equation (4) is a time term, and the second term on the left side is an advection term.

In the present embodiment, the above partial differential equation is solved by applying a discrete solution method by a CIP method (Cubic Interpolation Profile, see Non-Patent Document 8). Here, for convenience of explanation, the formula (4) will be described by replacing it with one dimension as the following formula.

Here, f is a one-dimensional quantity and corresponds to the image intensity. u is a velocity field vector. The image intensity of a certain pixel point i _f i, the differential value of _{f i} described as _{g i} = ∂f _i / ∂x. Here, a cubic interpolation function between two pixels is defined as follows.

Let F _i (x) be a function between [i−1, i]. The value of F _i (x) and its differential value are set as follows.

From the above relationship, the following equation is obtained.

The image at time n + 1 is obtained by advancing the image at time n by u · Δt, and the following equation is obtained.

In this way, the image f ^{n + 1 at the} time ^{n + 1} is obtained from the image at the time n by the CIP method.

Next, a process for obtaining the partial differential equation by the difference method will be described as a comparative example. Here, equation (4) is modified as follows.

  In the calculation method, Equation (17) is discretized by the difference method, and the calculation is recursively repeated along the two-dimensional image plane and the discrete time axis. The initial value of the image intensity I is one image input to the input unit 11. By repeating the calculation corresponding to the predetermined time, an image moved as much as necessary can be obtained.

  The display unit 16 displays the animated image by various methods. FIG. 6 is a diagram illustrating an example in which an image of a single tree is converted into an animation using speed field information obtained by learning, and FIG. 7 is a diagram illustrating a human lip using speed field information obtained by learning. It is a figure which shows the example which animated the image of no. FIG. 8 is a diagram illustrating an example in which a wave image is converted into a moving image using velocity field information obtained by learning.

  Each image has a coefficient learned in advance using a time-series image of another tree in FIG. 6, another time-series image of lips in FIG. 7, and another time-series image of waves in FIG. The velocity field vector obtained using the coefficient corresponding to the image most similar to the generated image is applied. However, since there are various directions and magnitudes of movement of the object, it is possible to linearly affine transform the velocity field vector and determine and reflect the affine matrix elements by trial and error according to the input image. desirable.

  FIG. 9 applies a wind velocity field vector obtained by learning to a still image in which the tree 900 and snow 910 are represented, and dynamically expresses how the snow 910 is blown by the wind while shaking the tree 900. This is an example.

  Next, the effect of improving the image quality when the CIP method of the embodiment is applied will be described by comparison with the difference method of the comparative example. FIG. 10 is a diagram showing the state of image generation according to the embodiment. In FIG. 10, the broken line represents the input original image profile, the solid line represents the ideal generated image profile, and the black circle represents the advection equation by the CIP method. This is a profile of the generated image. The profile of the original image represents a rectangular wave and a triangular wave, and a velocity field vector that moves to the right is given to the velocity component of the advection equation. The square wave was used to examine the preservation of the edge portion of the image, and the triangular wave was used to examine the preservation of the maximum value. FIG. 11 is a diagram showing a state of image generation according to the comparative example, and a black circle in the same figure is an image profile generated by solving the advection equation by the difference method. The broken line and the solid line in the figure are the same as those in FIG. In the difference method, the result when the recursive calculation is performed 1000 times is shown.

  Comparing FIG. 10 and FIG. 11, in the difference method, as shown in the broken-line circle in FIG. 11, the edge portion of the rectangular wave is disturbed by error diffusion. On the other hand, in the CIP method, as shown in the dotted circle in FIG. 10, such disturbance does not occur, and each waveform moves while maintaining the initial profile almost as it is. confirmed. This indicates that the CIP method is highly effective in suppressing unnatural disturbance of the image.

  FIG. 12 is an example of a screen displayed by the display unit 16 when the moving image generating apparatus 1 is applied to a graphical user interface (GUI). As shown in the figure, the display unit 16 displays an image 200 representing a tree 210 and a star 220, an icon 230 indicating a star, an icon 240 indicating the sun, and an icon 250 indicating light emission. The icon 230 corresponds to speed field information indicating the movement of the star obtained using an image similar to the star 220, the icon 240 is similar to the speed field information indicating the movement of the sun, and the icon 250 is similar to the star 220. Respectively corresponding to velocity field information indicating the movement of light radiation obtained using the image to be transmitted. When the user designates the star 220 in the screen using the mouse and designates the icon 230, the moving image generating apparatus 1 reads the velocity field information corresponding to the icon 230 from the storage device and substitutes it into the advection equation. A moving image showing the movement of 220 is generated and displayed on the screen. When the user designates the star 250 in the screen using the mouse and designates the icon 250, the velocity field information corresponding to the icon 250 is read from the storage device and substituted into the advection equation. A moving image showing light emission is generated and displayed on the screen. As described above, it is possible to generate a moving image by selecting a related movement from a menu for one image.

  Therefore, according to the present embodiment, the time series image and the velocity field information obtained using the image are stored in the storage device 17 in association with each other, and a moving image is generated from the image input by the input unit 11. When generating, the speed field search unit 14 searches for a time-series image similar to the input image, and the moving image generation unit 15 has the time field and the advection term for the speed field information corresponding to the similar image. By substituting and processing the advection term of the partial differential equation, the partial differential equation is processed using velocity field information suitable for the input image, so that a moving image with a sense of reality can be generated.

  According to the present embodiment, the velocity field information detected by applying the optical flow method to the time-series image by the velocity field detection unit 12 is used as velocity field information when generating a moving image of the input image. Thus, a moving image with simple movement can be easily generated.

  According to the present embodiment, the velocity field learning unit 13 learns the coefficient by substituting the time-series image and the velocity field information into the input / output function that outputs the velocity field information with respect to the input of the image and the coefficient. , By substituting the input image and this coefficient into the input / output function, calculate more appropriate velocity field information for the input image, and generate the video by substituting this velocity field information into the partial differential equation Thus, it is possible to generate an appropriate moving image even for an image of an object whose movement is complex.

  According to the present embodiment, the moving image generation unit 15 applies the discrete solution method by the CIP method to solve the partial differential equation, and the values of the adjacent pixels are interpolated by the cubic interpolation function. Since the image intensity is stable even if the speed change is large, numerical errors can be suppressed. In addition, since the time interval can be increased, a high speed can be realized.

  According to the present embodiment, by applying the moving image generating apparatus 1 to the GUI, it is possible to generate a moving image by selecting a related motion from a menu for one image, and even a beginner can feel a sense of reality. A moving image showing a complicated movement can be generated from a single image in a short time, and the production efficiency of moving image generation can be improved.

  In addition, as an image input by the input unit 11, various images can be used in addition to a real image, for example, an artificial image generated by computer graphics may be used.

It is a block diagram which shows the structure of the moving image production | generation apparatus in one embodiment. It is a figure which shows a mode that an optical flow method is applied to the time-sequential image of a wave, and the state which calculates | requires velocity field information, the figure (a) is an image of the time n, (b) is an image of the time n + 1, (c) is ( The velocity vector as velocity field information obtained from the images of a) and (b) is shown. It is a figure which shows a mode that an optical flow method is applied to a star-shaped image, and a state is calculated | required, the same figure (a) is a star-shaped image of time n, (b) is a star-shaped image of time n + 1, (c ) Indicates a velocity vector as velocity field information obtained from the images (a) and (b). It is a figure for demonstrating the process which searches the image similar to the input image, and reads the speed field information corresponding to the searched similar image. It is a figure which shows the speed field information calculated | required by applying the optical flow method as learning object. It is a figure which shows the example of the moving image produced | generated from the image of a tree. It is a figure which shows the example of the moving image produced | generated from the image of the lips. It is a figure which shows the example of the moving image produced | generated from the image of the river. It is a figure which shows the example which gave the fluctuation to the tree with respect to the image of a tree and snow, and moved snow with the wind. It is a figure which shows the mode of the image generation by an Example, The broken line of the figure is the profile of the input original image, A solid line is the profile of an ideal production | generation image, A black circle is the image produced | generated by solving an advection equation by CIP method It is a profile. It is a figure which shows the mode of the image generation by a comparative example, and the broken line of the figure is the profile of the input original image, the solid line is the profile of the ideal generated image, and the black circle is the image generated by solving the advection equation by the difference method It is a profile. It is a figure which shows the example of the display screen at the time of applying this moving image production | generation apparatus to GUI.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Movie production | generation apparatus 11 ... Input part 12 ... Speed field detection part 13 ... Speed field learning part 14 ... Speed field search part 15 ... Movie production | generation part 16 ... Display part 17 ... Memory | storage device

Claims (8)

Storage means for storing time-series images and velocity field information obtained using the images in association with each other;
An input means for receiving image input;
A search unit that searches the storage unit for an image similar to the input image, and reads speed field information corresponding to the searched image from the storage unit;
Generating means for generating a moving image by substituting and processing the input image and the read velocity field information into the advection term of the partial differential equation having a time term and an advection term;
A moving image generating apparatus comprising:   The moving image according to claim 1, wherein the advection term is expressed by a product of a velocity component and a spatial first derivative of image intensity, and the velocity field information is substituted into the velocity component. Generator.   3. The search unit according to claim 1, wherein the search unit calculates a similarity between the input image and a time-series image using a correlation function, and searches for a similar image based on the similarity. Video generator.   A velocity field that reads a time-series image from the storage means, detects velocity field information by applying an optical flow method to the image, and stores the velocity field information in the storage means in association with the time-series image. The moving image generating apparatus according to claim 1, further comprising a detecting unit. The time series image and the corresponding velocity field information are read from the storage means, and the coefficients are learned by substituting them into an input / output function that outputs velocity field information in response to the input of the image and the coefficient. The image and the coefficient are associated with each other and stored in the storage means,
A coefficient corresponding to a similar image searched by the search means is read from the storage means;
Learning means for substituting the input image and the read coefficient into the input / output function to calculate velocity field information and storing it in a storage means;
The moving image generating apparatus according to claim 1, wherein the search unit reads the velocity field information calculated by the learning unit from the storage unit.   6. The moving image generating apparatus according to claim 1, wherein the moving image generating means solves the partial differential equation by applying a discrete solution method using a CIP method. Storing a time-series image and speed field information obtained using the image in association with each other in a storage unit;
Receiving an image input by an input means;
Searching the storage means for an image similar to the input image by the search means, and reading speed field information corresponding to the searched image from the storage means;
A step of generating an animation by substituting and processing the input image and the read velocity field information into the advection term of the partial differential equation having a time term and an advection term by the generation means;
A moving image generating method characterized by comprising: A process of storing the time-series image and the velocity field information obtained by using the image in a storage unit in association with each other;
Input processing that accepts input of images;
A search process for searching for an image similar to the input image from the storage unit, and reading speed field information corresponding to the searched image from the storage unit;
Generation processing for generating a moving image by substituting the input image and the read velocity field information into the advection term of the partial differential equation having the time term and the advection term, and processing,
A moving picture generating program characterized by causing a computer to execute.

Images