IMAGE PROCESSING METHOD, APPARATUS AND MEMORY MEDIUM THEREFOR

Abstract

There is provided an image combining apparatus comprising an image input unit for entering photographed images, an image memory for temporarily storing the entered images, a corresponding point extracting unit for extracting the corresponding point information between the images stored in the image memory, a photographing method discriminating unit for discriminating the photographing method at the photographing operation based on the extracted corresponding point information, a combination process unit for appropriately combining the images according to the discriminated photographing method, and a combined image memory for storing the combined image. The corresponding point extraction unit extracts the corresponding points between two images, and the photographing method discrimination unit discriminates whether the translational photographing method or the panning photographing method is used. The combination process unit executes combination of the images, based on the discriminated photographing method.

Description

BACKGROUND OF THE INVENTION

[0001] 1. Field of the Invention

[0002] The present invention relates to an image combining method for combining partially overlapping plural images to generating a panoramic image or a high definition image, and an apparatus and a memory medium therefor.

[0003] 2. Related Background Art

[0004] In case of generating for example a panoramic image by combining plural images, it is important to select an appropriate combining process according to the photographing method. The combining process mentioned above is for example an affin conversion, a projection conversion, a cylindrical mapping conversion etc.

[0005] Conventionally, the photographing method is either entered by the operator at the combining operation, or estimated by the output information of a hardware such as gyro, mounted on the camera at the photographing operation.

[0006] In the former method, however, the operator is required to enter the photographing method at the combining operation and has therefore to remember the situation of the photographing operation. On the other hand, the latter method is associated with a drawback that the camera itself becomes bulky since a detector such as a gyro has to be mounted on the camera. Also, in case of an off-line combining operation, the output of the gyro has to be retained by a suitable method.

SUMMARY OF THE INVENTION

[0007] In consideration of the foregoing, an object of the present invention is to provide an image processing method utilizing images and capable of automatically selecting the combining method.

[0008] Another object of the present invention is to use an image for discriminating the photographing method and to effect image combination according to the photographing method, thereby achieving image combination of a higher definition, depending on the photographing method.

[0009] Still another object of the present invention is to evaluate the image combination based on at least one of plural combining methods, thereby selecting an appropriate combining method.

[0010] Still another object of the present invention is to extract a corresponding point in an overlapping area of plural images, thereby selecting the combining method in more exact manner.

[0011] Still another object of the present invention is to enable easier discrimination of translation displaced photographs or panned photographs by employing cylindrical mapping conversion or spherical mapping conversion.

[0012] Still another object of the present invention is to enable automatic formation of a panoramic (for 360° around) image with a higher precision.

[0013] An image processing method according to an aspect of the present invention comprises an input step of entering plural images, and a selection step of selecting an appropriate combining method from plural combining methods, utilizing the plural image information.

[0014] Still other objects of the present invention, and the features thereof, will become fully apparent from the following detailed description of the embodiments, to be taken in conjunction with the attached drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

[0015] FIG. 1 is a block diagram showing the configuration of an image combining apparatus of a first embodiment;

[0016] FIG. 2 is a view showing a translation photographing mode in case of photographing for example an original image;

[0017] FIG. 3 is a view showing a panning photographing mode in case of photographing for example a landscape;

[0018] FIG. 4 is a flow chart showing an image combining sequence of the image combining apparatus;

[0019] FIG. 5 is a view schematically showing a corresponding point extracting process;

[0020] FIGS. 6A and 6B are views schematically showing corresponding points in images obtained in translation photographing and panning photographing;

[0021] FIG. 7 is a flow chart showing a photographing method discriminating sequence on a photographing method discriminating unit 102;

[0022] FIGS. 8A and 8B are views showing a combination process by developing a combination image memory 105;

[0023] FIG. 9 is a block diagram showing the configuration of an image combination apparatus of a second embodiment;

[0024] FIG. 10 is a view showing a cylindrical mapping conversion process in an image conversion unit 109;

[0025] FIG. 11 is a view schematically showing an image after conversion;

[0026] FIG. 12 is a flow chart showing a photographing method discriminating sequence in a photographing method discrimination unit 107;

[0027] FIG. 13 is a view showing a combined image in case panning photographing is identified;

[0028] FIG. 14 is a block diagram showing an image combination apparatus of a third embodiment;

[0029] FIGS. 15 and 16 are flow charts showing an image combining sequence in the third embodiment;

[0030] FIG. 17 is a view showing n combined images after parameter conversion;

[0031] FIG. 18 is a view showing the viewing angles in the respective images;

[0032] FIG. 19 is a view showing a center line S of the overlapping area of images;

[0033] FIG. 20 is a view showing combined images prior to conversion process;

[0034] FIG. 21 is a view showing combined images after conversion process;

[0035] FIG. 22 is a view schematically showing the conversion process;

[0036] FIG. 23 is a view schematically showing generated combined images;

[0037] FIG. 24 is a view showing generation of an image by 360° cylindrical mapping;

[0038] FIG. 25 is a block diagram of a computer system realizing an image combining apparatus of a fourth embodiment; and

[0039] FIG. 26 is a flow chart showing an image combining sequence in the fourth embodiment.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0040] FIG. 1 is a block diagram showing the configuration of an image combination apparatus of a first embodiment of the present invention, wherein shown are an image combination process unit 1, an image input unit 100 for entering photographed images, an image memory 104 for temporarily storing the input images, and a corresponding point extraction unit 101 for extracting the corresponding point information among the images stored in the image memory 104.

[0041] There are further provided a photographing method discrimination unit 102 for discriminating the photographing method at the photographing operation, based on the information of the extracted corresponding points, a combination process unit 103 for appropriately combining the images according to the discriminated photographing method, and a combined image memory 105 for storing the combined images.

[0042] The photographing method discriminated by the photographing method discrimination unit 102 includes, for example, translation photographing in which the photographs are taken by varying the photographing position without changing the photographing direction, and panning photographing in which the photographs are taken by varying the photographing direction without changing the photographing position. FIG. 2 shows translation photographing in case of photographing for example original images. After the photographing at a position 1, the camera is moved to a position 2 for photographing again. The movement from the position 1 to 2 involves changes in the main parameters of translation movement (Δx, Δy), rotation (θ) and magnification difference m. Such photographing mode will be called translation photographing.

[0043] FIG. 3 shows the panning photographing in case of photographing for example a landscape. The photographing operations are executed with the camera in positions 1, 2 and 3. Main changes involve rotation (ψ, φ, θ) about the X, Y, Z axes. In particular, the rotation φ about the Y axis is a governing factor since the photographing operations are executed in succession while the camera is rotated in the horizontal direction. Such photographing mode will be called panning photographing.

[0044] The photographing method discrimination unit 102 discriminates, in the present embodiment, whether the photographing method is translation photographing or panning photographing.

[0045] In the following there will be explained the function of the image combining apparatus of the above-described configuration. FIG. 4 is a flow chart showing the image combining sequence of the image combining apparatus. At first the image input unit 100 enters image data (step S1). The input image data consists of an image photographed with a camera and digitized by A/D conversion, but there may also be entered an image stored in a recording medium or a memory device.

[0046] The image entered by the image input unit 100 is temporarily stored in the image memory 104 (step S2). Then there is discriminated whether two images have been entered (step S3). If two images have been entered, the entered two images are used for executing a process of extracting corresponding points (step S4). FIG. 5 schematically shows the corresponding point extracting process. The corresponding points are extracted in the following manner, utilizing the images 200L, 200R in the image memory 104:

[0047] 1) There is set a template 201 of p×p pixels, with the center at a point P1L on a reference image 200L.

[0048] 2) A correlation value Zmn is generated according to the following equation (1) at each moved point, while the template 201 is moved within a search area 202 set on a reference image 200R: 1$\begin{array}{cc}{z}_{mn}=\frac{\sum _x\sum _y{f}_L\left(x,y\right)·f_R\left(x-m,y-n\right)}{\sqrt{\sum _x\sum _y{f_L\left(x,y\right)}^2}\sqrt{\sum _x\sum _y{f_R\left(x-m,y-n\right)}^2}}& \left(1\right)\end{array}$

[0049] wherein ΣxΣy indicates summation on all the pixels in the template.

[0050] 3) Correlation values Zmn on the moved points on the reference image 200R are compared to extract a point providing the maximum value.

[0051] 4) For the point P1L (i, j), the coordinate (i′, j′) of the corresponding point is given by i′=i−m′, j′=j−n′ wherein (m′, n′) indicates the movement amount providing the maximum correlation value.

[0052] The corresponding point extraction unit 101 sets plural templates on the image 200L shown in FIG. 5 and extracts the corresponding point for each template, according to the procedure 1) to 4) explained above. The extracted corresponding points are stored in an unrepresented memory.

[0053] Based on the corresponding point information extracted in the step S4, the photographing method discrimination unit 102 discriminates the photographing method (step S5). In the following there will be explained the discrimination of the photographing method.

[0054] FIGS. 6A and 6B schematically show the corresponding points between the images obtained by translation photographing and panning photographing. The coordinate relationship between the corresponding points shown in FIG. 6A can be described by affin conversion represented by the equations (2):

x′=Ax+By+C

y′=−Bx+Ay+D   (2)

[0055] wherein (x, y) is the coordinate on the image 300L in FIG. 6A, while (x′, y′) is the coordinate on the image 300R. Also the coordinate relationship between the corresponding points shown in FIG. 6B can be described by projection conversion represented by the following equations (3): 2$\begin{array}{cc}{x}^{\prime }=\frac{m_{1}x+m_{2}y+m_3}{m_{7}x+m_{8}y+1}{y}^{\prime }=\frac{m_{4}x+m_{5}y+m_6}{m_{7}x+m_{8}y+1}& \left(3\right)\end{array}$

[0056] FIG. 7 is a flow chart showing the photographing method discriminating sequence in the photographing method discrimination unit 102. At first the corresponding point information obtained in the corresponding point extraction unit 101 is read from a memory which is not shown (step S11).

[0057] The read corresponding points are subjected to calculations by the least square method, based on the conversions represented by the foregoing equations (2) and (3), for calculating a combination error for evaluating the quality of image combination (step S12). The evaluating equation by the least square method is shown in the following. In case of the affin conversion, there are calculated parameters A, B, C and D for minimizing an error E2-of the equation (4):

E 2=Σ[{ x′− (Ax+By+C)}2+{y′−(−Bx+Ay+D)}2]  (4)

[0058] Also in case of the projection conversion, there are calculated parameters m1 to m8 for minimizing an error E3 of the following equation (5): 3$\begin{array}{cc}\begin{array}{c}\mathrm{E3}=\sum [{\left\{x^{\prime }\left(m_{7}x+m_{8}y+1\right)-\left(m_{1}x+m_{2}y+m_3\right)\right\}}^2+\\ {\left\{y^{\prime }\left(m_{7}x+m_{8}y+1\right)-\left(m_{4}x+m_{5}y+m_6\right)\right\}}^2]\end{array}\&AutoLeftMatch;& \left(5\right)\end{array}$

[0059] The error E2 obtained in the step S12 is compared according to a relation (6) (step S13):

E 2<Th  (6)

[0060] Thus, if the error E2 is smaller than a predetermined threshold value Th, there is identified the affin conversion, namely the translation photographing method (step S15). On the other hand, if the error E2 is at least equal to the predetermined threshold value Th, the error E2 is compared with the error E3 according to a relation (7):

E 2≦E3  (7)

[0061] Thus, if the error E2 does not exceed the error E3, there is identified the affin conversion or namely the translation photographing (step S15), but, if the error E3 is smaller than the error E2, there is identified the projection conversion, namely the panning photographing (step S16). The result of discrimination is given, together with the parameters, to the combination process unit 103.

[0062] Then a combination process is executed, based on the result of discrimination in the step S5 (step S6). FIGS. 8A and 8B illustrate a combination process, showing the development of a combination image memory 105. FIG. 8A shows the combination process in case the translation photographing is discriminated, while FIG. 8B shows that in case the panning photographing is discriminated.

[0063] The combination process unit 103 executes the combination process in the following sequence:

[0064] 1) A joint position S between the images is set, based on the combination parameters.

[0065] 2) The image 200L is read from the image memory 104 and the area at the left side of the joint S is written into the combination image memory 105.

[0066] 3) The area at the right side of the joint S is written into the combination image memory 105, based on the image 200R. The coordinate (x′, y′) on the image 200R, corresponding to a point P1 (x, y) in FIG. 8A is calculated by the equations (2). Also the coordinate on the image 200R, corresponding to a point P2 (x, y) in FIG. 8B is calculated by the equations (3).

[0067] 4) The pixel value at the corresponding coordinate (x′, y′) on the image 200R is written as the pixel value at the point P1 (x, y) or P2 (x, y). The combination process is executed in this manner.

[0068] In case the corresponding coordinate value (x′, y′) in the foregoing sequence step 3) contains a decimal fraction, the pixel value may be generated by a known interpolating process.

[0069] Through the sequence S1 to S6 in FIG. 4, the combined image is stored in the combination image memory 105 as shown in FIGS. 8A and 8B.

[0070] As the image combination apparatus of the first embodiment discriminates the photographing method solely from the images and executes the optimum combination process according to the result of discrimination, the user need not remember the photographing method and a high-quality combined image can be constantly obtained.

[0071] Even when the combination method does not correspond to the photographing method, the image combination is executed by such combination method if the combination error is sufficiently small, and the obtained result is completely acceptable since the appropriate combination method is selected also in such case.

[0072] Also in the generation of the combination parameters etc., it is also possible to calculate the correlation value of the overlapping area by displacing an image by a pixel at a time, or deforming an image, with respect to the other image.

[0073] The process time can be reduced by extracting the corresponding points and utilizing the information of such corresponding points, since the calculation becomes simpler.

[0074] FIG. 9 is a block diagram of an image combination apparatus of a second embodiment of the present invention, wherein components same as those in the first embodiment are referred to by same numbers and will not be explained further.

[0075] The image combination apparatus of the second embodiment is featured by a focal length input unit 106, an image conversion unit 109, a photographing method discrimination unit 107 and a combination process unit 108.

[0076] The focal length input unit 106 serves to enter the focal length at the photographing operation. The focal length may be entered by the user either by the input of a numerical value through a keyboard or by the selection from plural values, or by recording the focal length information together with the image at the photographing operation and reading the recorded information by the focal length input unit 106.

[0077] The image conversion unit 109 converts the image data utilizing the focal length, based on the cylindrical mapping conversion. Corresponding point extraction units 1010, 1011 extract the corresponding points between the images, according to a process similar to that in the first embodiment.

[0078] The photographing method discrimination unit 107 discriminates, as in the first embodiment, whether the photographing method is translation photographing or panning photographing, but employed evaluation method is different from that of the first embodiment. The combination process unit 108 executes image combination according to the discriminated photographing method.

[0079] In the following there will be explained the function of the image combination apparatus of the second embodiment, having the above-described configuration, wherein steps similar to those in the foregoing first embodiment are numbered same and will not be explained further.

[0080] The image conversion unit 109 executes cylindrical mapping conversion on the image data stored in the image memory 104, based on the focal length obtained by the focal length input unit 106. FIG. 10 shows the cylindrical mapping conversion process in the image conversion unit 109.

[0081] In FIG. 10, there are shown an original image 200 read from the image memory 104, and an image 201 after the cylindrical mapping conversion. The coordinate (x, y) before the cylindrical mapping is performed and the coordinate (φ, v) after the cylindrical mapping is performed are correlated by the following equations (8): 4$\begin{array}{cc}x=ftan\phi y=\frac{\sqrt{x^2+f^2}}{f}v& \left(8\right)\end{array}$

[0082] The image conversion unit 109 converts the image according to the equations (8). FIG. 11 schematically shows the image after conversion. In the image combination process unit 2 of the image combination apparatus of the second embodiment, the corresponding point extraction unit 1010 executes extraction of the corresponding points between the original images, while the corresponding point extraction unit 1011 executes extraction of the corresponding points utilizing the images converted by the cylindrical mapping conversion. Consequently there are generated two sets of paired corresponding points.

[0083] The photographing method discrimination unit 107 discriminates the photographing method, utilizing the corresponding point information generated by the corresponding point extraction units 1010, 1011.

[0084] FIG. 12 is a flow chart showing the photographing method discrimination sequence in the photographing method discrimination unit 107. At first the corresponding point information obtained in the corresponding point extraction units 1010, 1011 is read from a memory which is not shown (step S21).

[0085] Thus read corresponding points are used in the calculations by the least square method, based on the foregoing conversion equations (2) (step S22).

[0086] The equations (2) alone are employed in order to effect affin coordinate conversion on the image obtained by cylindrical mapping conversion, even in case of the image taken by the panning photographing. Stated differently, an appropriate combination method can be selected by calculating the combination error by the equations (2) on the image subjected to cylindrical mapping conversion and the image not subjected to such cylindrical mapping conversion.

[0087] Equations (9) show the evaluation by the least square method. More specifically, there are calculated parameters Ao, Bo, Co, Do, As, Bs, Cs, Ds minimizing the errors E4, E5 of the equations (9), on the paired corresponding points (xo, yo)−(xo′, yo′) obtained by the extraction from the original images and those (xs, ys)−(xs′, ys′) obtained by the extraction from the images subjected to cylindrical mapping conversion: 5$\begin{array}{cc}\begin{array}{c}\mathrm{E4}=\sum [{\left\{x_O^{\prime }-\left(A_O{x}_O+B_O{y}_O+C_O\right)\right\}}^2+\\ {\left\{y_O^{\prime }-\left(-B_O{x}_O+A_O{y}_O+D_O\right)\right\}}^2]\\ \mathrm{E5}=\sum [{\left\{x_S^{\prime }-\left(A_S{x}_S+B_S{y}_S{C}_S\right)\right\}}^2+\\ {\left\{y_S^{\prime }-\left(-B_S{x}_S+A_S{y}_S+D\right)\right\}}^2]\end{array}\&AutoLeftMatch;& \left(9\right)\end{array}$

[0088] Then comparison is made whether the error E4 obtained in the step S22 is smaller than a threshold value Th (E4<Th) (step S23). If the error E4 is smaller than the predetermined threshold value Th, there is identified the translation photographing and there is selected a combination method corresponding to the translation photographing (step S25).

[0089] If the error E4 is at least equal to the threshold value Th, comparison is made whether the error E4 exceeds the error E5 (E4≦E5) (step S24). If the error E4 does not exceed the error E5, there is identified the translation photographing (step S25), but, if the error E5 is smaller than E4, there is identified the panning photographing (step S26), and, after the selection of a combination method corresponding to the result of such discrimination, the sequence is terminated. The result of discrimination is given, together with the parameters, to the combination process unit 108.

[0090] The function of the combination process unit 108 is basically same as that in the foregoing first embodiment. It is however different from the first embodiment in that, in case the panning photographing is identified, the photographing method discrimination unit 107 reads the original images from the image memory 104, then converts the images by cylindrical mapping conversion by a process similar to that in the image conversion unit 109 and writes the pixel values in the combined image memory 105 utilizing the above-mentioned parameters As, Bs, Cs, Ds.

[0091] FIG. 13 shows a combined image in case the panning photographing is identified. The cylindrical mapping conversion is executed again at the image combination, but it is also possible that the image conversion unit 109 stores the combined image in the image memory 104 and the combination process unit 108 executes the image combination by reading the stored image after cylindrical mapping conversion.

[0092] It is to be noted that a better image may be obtained by the projection conversion as in the first embodiment, if the image angle of the combined image obtained from the panning photographing is not too large.

[0093] In such case, the image combining method may be switched according to the image angle, after the discrimination of the panning photographing.

[0094] Also, in the panning photographing operation explained in the foregoing, it is assumed that the photographing direction is changed only in one direction (for example lateral direction), but, if the photographing direction is changed in two or more directions (for example lateral and vertical directions), the known spherical mapping conversion is more preferable and the image combining method may be switched also in such case.

[0095] FIG. 14 is a block diagram showing the configuration of an image combination apparatus of a third embodiment of the present invention, which is featured in executing a process corresponding to the combination of n images. The image combination process unit 1 in the third embodiment is similar in configuration to that in the first embodiment.

[0096] Referring to FIG. 14, an n-image integrating unit 111 generates and retains various information for combining n images. A focal length estimation unit 112 estimates the focal length. Consequently the third embodiment is applicable also in case the focal length is unknown. However, if the focal length is already known, input means may be separately provided for entering the value of the focal length. There are also provided a main image memory 113 for storing information of n images, and an n-image integrating unit 114 for generating a panoramic image by combining n images.

[0097] FIGS. 15 and 16 are flow charts showing an image combining sequence in the third embodiment. At first two images are read from the main image memory 113 and supplied into the image combination process unit 1 (step S31). It is assumed that the number of images is stored, by means not shown, in the n-image integrating unit 111.

[0098] The image combination process unit 1 combines two images in a process similar to that in the foregoing first embodiment (step S32). The combined image retained in the combined image memory 105 may be displayed in display unit not shown, in order that the user can confirmed the result of combination. The coordinate data of the corresponding points between the images, the combining parameters and the information on the photographing method are retained in the n-image integrating unit 111.

[0099] Then there is discriminated whether combination of n images has been completed (step S33). If not, the sequence returns to the step S31 to repeat the sequence, but, if completed, the results of (n−1) discriminations, retained in the n-image integrating unit 111, are checked and the photographing method for the n images is determined by majority as the translation photographing or the panning photographing (step S34).

[0100] The result of discrimination of the photographing method determined in the step S34 is identified (step S35). More specifically, if the translation photographing is discriminated, the information on the corresponding points among the images, retained in the n-image integrating unit 111, is read in succession and the affin conversion parameters Ai, Bi, Ci, Di are calculated by the least square method according to the equations (2) (step S36).

[0101] Then thus calculated affin conversion parameters are converted into combination parameters Ai′, Bi′, Ci′, Di′ based on the image 1 (step S37). FIG. 17 shows the combination of n images after the conversion of the combination parameters.

[0102] The n-image combination unit 114 uses the converted combination parameters Ai′, Bi′, Ci′, Di′ to generate the pixel value of each point based on the coordinate conversion of the equations (2) and stores such pixel values in the main image memory 113 (step S38). The method of generating the pixel value is similar to that in the foregoing first embodiment.

[0103] On the other hand, if the step S35 discriminates the panning photographing, there is displayed, on display unit not shown, a message for confirming a 360° combination (step S39), and there is discriminated whether 360° combination is instructed (step S40). If 360° combination is not instructed, the focal length estimation unit 112 estimates the focal length in the following manner (step S41).

[0104] As shown in FIG. 3, the panning photographing operation is executed by changing the viewing field by the rotation (ψ, φ, θ) about the X, Y, Z axes. Consequently, for converting the coordinate between the images, there is given a matrix M represented as (10):

M=FRF −1

[0105] wherein: 6$\begin{array}{cc}R=\left[\begin{array}{ccc}\cos \mathrm{\psi cos\phi }& \sin \mathrm{\psi cos\theta }+\cos \mathrm{\psi sin\phi sin\theta }& \sin \mathrm{\psi sin\theta }-\cos \mathrm{\psi sin\phi cos\theta }\\ -\sin \mathrm{\psi cos\phi }& \cos \mathrm{\psi cos\theta }-\sin \mathrm{\psi sin\phi sin\theta }& \cos \mathrm{\psi sin\theta }+\sin \mathrm{\psi sin\phi cos\theta }\\ \sin \phi & -\cos \mathrm{\phi sin\theta }& \cos \mathrm{\phi cos\theta }\end{array}\right]F=\left[\begin{array}{ccc}1& 0& 0\\ 0& 1& 0\\ 0& 0& 1/f\end{array}\right]& \left(10\right)\end{array}$

[0106] According to the equations (3), the coordinate conversion matrix can be represented as (11): 7$\begin{array}{cc}M=\left[\begin{array}{ccc}{m}_1& m_2& m_3\\ m_4& m_5& m_6\\ m_7& m_8& 1\end{array}\right]& \left(11\right)\end{array}$

[0107] From the equations (10) and (11), there is derived a relation represented by the equation (12): 8$\begin{array}{cc}f=\sqrt{\frac{m_1^2+m_4^2-m_2^2-m_5^2}{m_8^2-m_7^2}},m_8\ne m_{7}orf=\sqrt{\frac{-m_1{m}_2-m_4{m}_5}{m_6{m}_7}},m_6\ne 0,m_7\ne 0& \left(12\right)\end{array}$

[0108] Consequently, the focal length estimation unit 112 estimates the focal length by the equation (12), utilizing, as the parameters m1 to m7, those stored in the n-image integrating unit 111 corresponding to the panning photographing. As there are plural sets of parameters, there is employed the center value of the focal lengths fk obtained from such parameters. The estimated focal length f is retained in the n-image integrating unit 111.

[0109] The n-image combining unit 114 generates a combined image, utilizing the generated parameter f (step S42). As in the second embodiment, the combined image is subjected to a cylindrical mapping conversion to generate a combined image of n images as shown in FIG. 13, and the combined image is written into the main image memory 113.

[0110] On the other hand, if the step S40 identifies 360° combination, there is executed estimation of the coordinate conversion parameters between the first and n-th images, since they mutually overlap (step S43). This operation can be achieved by reading the first and n-th images from the main image memory 113 and executing the aforementioned process in the combined image process unit 1. The information such as the generated parameters is stored in the n-image integrating unit 111.

[0111] The focal length estimation unit 112 estimates the focal length (step S44). FIG. 18 shows the viewing angle ωk in each image, which is obtained from the following equation (13):

ωk=tan−1(xk/f)  (13)

[0112] wherein xk is obtained by determining the center line S of the overlapping area between the images, as shown in FIG. 19. The n-image integrating unit 111 determines xk by calculating the overlapping area between the images based on the coordinate conversion parameters.

[0113] Since the viewing angle of 360° is covered by n images, the sum of ω1 to ωn becomes 360° (2π), and this relationship is represented by the equation (14):

2π−Σ{(2 tan−1(xk/f)}=0  (14)

[0114] The focal length is estimated by solving the equation (14) for example by the Newton method. The estimated focal length is stored in the n-image integrating unit 111. Also the n-image integrating unit 111 executes cylindrical mapping conversion of the coordinates of the corresponding points between the images, based on the focal length f, and generates parameters according to the equations (2), utilizing thus converted coordinates of the corresponding points. The generated parameters are further converted into values based on a reference image (for example image 1) for use in the combination of all the images.

[0115] In this manner the focal length f can be automatically obtained, also in the second embodiment, solely from the image information.

[0116] Then there is executed a conversion process for providing a smooth joint between the first and n-th images as a result of 360° combination (step S45). FIG. 20 shows the combined image after such conversion process. Without the conversion process, the reference lines set respectively in the first and n-th images are mutually displaced.

[0117] FIG. 21 shows the combined image after such conversion process, wherein the reference lines respectively set in the first and n-th images mutually coincide. FIG. 22 schematically shows the conversion process.

[0118] As shown in FIG. 22, a start point P1 and an end point PN of the entire circumference are set, and the images are so converted that the reference lines of the images 1 and n mutually coincide at the point PN. In the illustrated case, the reference lines have a mutually aberration D at the point PN, while H pixels are present between the start point P1 and the end point PN, so that the write-in coordinate in the vertical direction is shifted by a value dν(x), given by the following equation (15), taking the start point P1 as the start point of the combined image in the horizontal direction: 9$\begin{array}{cc}dv\left(x\right)=\frac{D}{H}x& \left(15\right)\end{array}$

[0119] The n-image integrating unit 111 generates and retains the coefficient D/H in the equation (15), and the n-image combining unit 114 generates the combined image (step S46). In this operation, it generates the pixel values of the combined image by executing the cylindrical mapping conversion according to the equations (8) and the conversion according to the equation (15) and stores the generated pixel value in the main image memory 113.

[0120] The aberration in the image can thus be made inconspicuous, by distorting the entire image with a constant rate in the panning direction.

[0121] FIG. 23 schematically shows the combined image thus generated, which is displayed on display unit not shown. The above-described process allows to generate an automatically mapped cylindrical image free from aberration over the entire circumference of 360°. FIG. 24 shows the generation of the cylindrical image, mapped over the entire circumference of 360°.

[0122] FIG. 25 is a block diagram showing a computer system realizing an image combination apparatus of a fourth embodiment of the present invention. In the fourth embodiment, a program embodying the image combination process of the foregoing first embodiment is recorded on a recording medium and is rendered operable on the computer system.

[0123] In FIG. 25, there are shown a memory 120 for storing various data, a CPU 121, a reproduction unit 122 for reproducing the information of a recording medium, a display unit 123, and a recording medium 124.

[0124] FIG. 26 is a flow chart showing the image combination sequence in the fourth embodiment. A corresponding program is stored in the recording medium 124 and is executed by the CPU 121. At first the user selects images to be used in combination (step S51), and the user sets the arrangement of the images (step S52). Based on such arrangement, two neighboring images are read into the memory 120 (step S53).

[0125] From thus read images, corresponding points are extracted (step S54), and the photographing method is discriminated from the information of the corresponding points (step S55). Based on the discriminated photographing method, an optimum combining method is set and the combination parameters are generated (step S56).

[0126] Two images are combined utilizing the combination parameters (step S57), and the combined image is displayed (step S58). Then there is discriminated whether the combinations of the neighboring two images have been completed (step S59).

[0127] If not, the sequence returns to the step S53 to repeat the combination of two images, but, if completed, the parameters are converted into values relative to the reference image. All the images are combined, based on thus converted combination parameters (step S61). The process in each step is similar to that explained in the foregoing embodiments.

[0128] The recording medium can be composed not only of a memory element such as ROM but also of any other memory device such as a floppy disk, a hard disk, an optical disk, a magnetooptical disk, a CD-ROM, a CD-R, a DVD, a magnetic tape or a non-volatile memory card.

[0129] Many widely different embodiments of the present invention may be constructed without departing from the spirit and scope of the present invention. It should be understood that the present invention is not limited to the specific embodiments described in the specification, except as defined in the appended claims.

Claims

1. An image processing method comprising: an input step of entering plural images; and a selection step of selecting an appropriate image combining method among plural image combining methods, utilizing said plural images.

2. An image processing method according to claim 1, further comprising: an evaluation step of evaluating image combination based on at least one of said plural image combining methods; wherein said selection step selects the image combining method according to the result of said evaluation.

3. An image processing method according to claim 2, further comprising: an extraction step of extracting corresponding points on an overlapping area of said plural images.

4. An image processing method according to claim 3, wherein said evaluation step is adapted to calculate a combination error generated in case of image combination based on said corresponding points and wherein said selection step is adapted to evaluate the image combining method by comparing said combination error with a predetermined value.

5. An image processing method according to claim 3, wherein said evaluation step is adapted to calculate combination errors generated in case of image combinations respectively by at least two image combining methods based on said corresponding points and wherein said selection step is adapted to evaluate the image combining method by comparing the magnitude of said combination errors.

6. An image processing method according to claim 1, wherein said plural image combining methods at least include those corresponding to a translation photographing method in which images are taken with a change in the photographing position but without a change in the photographing direction, and a panning photographing method in which images are taken with a change in the photographing direction but with a change in the photographing position.

7. An image processing method according to claim 6, wherein the image combining method corresponding to said panning photographing method utilizes images obtained by coordinate conversion from said plural images onto a virtual cylindrical plane or a virtual spherical plane having a radius equal to the photographing focal length.

8. An image processing method according to claim 6, wherein the image combining method corresponding to said panning photographing method includes a correction step adapted, in case of generating a 360° panoramic image, to correct the image aberration generated in the 360° panoramic image generated by combining the both ends of the combined image, by distorting the entity of said combined image.

9. A recording medium storing a computer readable program for causing a computer to execute a process comprising: an input step of entering plural images; and a selection step of selecting an appropriate image combining method among plural image combining methods, utilizing said plural images.

10. A recording medium according to claim 9, wherein said process further comprises: an evaluation step of evaluating image combination based on at least one of said plural image combining methods; wherein said selection step selects the image combining method according to the result of said evaluation.

11. A recording medium according to claim 10, wherein said process further comprises an extraction step of extracting corresponding points on an overlapping area of said plural images.

12. A recording medium according to claim 11, wherein said evaluation step is adapted to calculate a combination error generated in case of image combination based on said corresponding points and wherein said selection step is adapted to evaluate the image combining method by comparing said combination error with a predetermined value.

13. A recording medium according to claim 11, wherein said evaluation step is adapted to calculate combination errors generated in case of image combinations respectively by at least two image combining methods based on said corresponding points and wherein said selection step is adapted to evaluate the image combining method by comparing the magnitude of said combination errors.

14. A recording medium according to claim 9, wherein said plural image combining methods at least include those corresponding to a translation photographing method in which images are taken with a change in the photographing position but without a change in the photographing direction, and a panning photographing method in which images are taken with a change in the photographing direction but with a change in the photographing position.

15. A recording medium according to claim 14, wherein the image combining method corresponding to said panning photographing method utilizes images obtained by coordinate conversion from said plural images onto a virtual cylindrical plane or a virtual spherical plane having a radius equal to the photographing focal length.

16. A recording medium according to claim 14, wherein the image combining method corresponding to said panning photographing method includes a correction step adapted, in case of generating a 360° panoramic image, to correct the image aberration generated in the 360° panoramic image generated by combining the both ends of the combined image, by distorting the entity of said combined image.

17. An image processing apparatus comprising: input means for entering plural images; and selection means for selecting an appropriate image combining method among plural image combining methods, utilizing said plural images.

18. An image processing apparatus according to claim 17, further comprising: evaluation means for evaluating image combination based on at least one of said plural image combining methods; wherein said selection means selects the image combining method according to the result of said evaluation.

19. An image processing apparatus according to claim 18, further comprising: extraction means for extracting corresponding points on an overlapping area of said plural images.

20. An image processing apparatus according to claim 19, wherein said evaluation means is adapted to calculate a combination error generated in case of image combination based on said corresponding points and wherein said selection means is adapted to evaluate the image combining method by comparing said combination error with a predetermined value.

21. An image processing apparatus according to claim 19, wherein said evaluation means is adapted to calculate combination errors generated in case of image combinations respectively by at least two image combining methods based on said corresponding points and wherein said selection means is adapted to evaluate the image combining method by comparing the magnitude of said combination errors.

22. An image processing apparatus according to claim 17, wherein said plural image combining methods at least include those corresponding to a translation photographing method in which images are taken with a change in the photographing position but without a change in the photographing direction, and a panning photographing method in which images are taken with a change in the photographing direction but with a change in the photographing position.

23. An image processing apparatus according to claim 22, wherein the image combining method corresponding to said panning photographing method utilizes images obtained by coordinate conversion from said plural images onto a virtual cylindrical plane or a virtual spherical plane having a radius equal to the photographing focal length.

24. An image processing apparatus according to claim 22, wherein the image combining method corresponding to said panning photographing method includes correction means adapted, in case of generating a 360° panoramic image, to correct the image aberration generated in the 360° panoramic image generated by combining the both ends of the combined image, by distorting the entity of said combined image.

25. An image processing method comprising: an input step of entering plural images; an extraction step of extracting corresponding points in an overlapping area of said plural images; and a selection step of selecting an image combining method based on said corresponding points.

26. An image processing method according to claim 25, further comprising: an evaluation step of evaluating image combination based on at least one of said plural image combining methods; wherein said selection step selects the image combining method according to the result of said evaluation.

27. An image processing method according to claim 25, wherein said plural image combining methods at least include those corresponding to a translation photographing method in which images are taken with a change in the photographing position but without a change in the photographing direction, and a panning photographing method in which images are taken with a change in the photographing direction but with a change in the photographing position.

28. An image processing method according to claim 26, wherein said evaluation step cylindrical mapping conversion or spherical mapping conversion.

29. An image processing method according to claim 25, wherein said image combining method utilizes affin conversion or projection conversion.

30. An image processing method according to claim 25, further comprising a correction step adapted, in case of generating a 360° panoramic image, to correct the image aberration generated in the 360° panoramic image generated by combining the both ends of the combined image, by distorting the entity of said combined image.