TWI590658B - Image fusion method for multiple lens and device thereof - Google Patents

Description

Multi-lens image fusion method and device thereof

The invention provides an image fusion method and a device thereof, in particular to an image fusion method and a device using the same.

The current image fusion method is mainly a pyramid fusion algorithm, and other image fusion methods are, for example, SWT, CVT, NSCT, GRW, WSSM, HOSVD, GFF, and the like. These image fusion methods combine multiple images of the same view but different focal lengths into a new image, and this new image is composed of the clearest parts of each image. Therefore, the camera needs to take multiple images with the same view but different focal lengths at different times.

If the camera moves during the shooting process, it will fuse the wrong result. In addition, there are time lags in multiple images taken at different times, resulting in scenes in each image that are slightly different, such as vehicle movement or leaf swaying, resulting in artifacts or blurring of the image fusion.

Therefore, in the process of shooting, if the camera movement can be reduced and the time difference of each image at different times can be avoided at the same time, the result of image fusion can be improved.

The invention provides an image fusion method using multi-lens and a device thereof, which simultaneously captures a plurality of parallax images with different focal lengths through multiple lenses, and adjusts more Parallax images for image fusion. Accordingly, the multi-lens image fusion method and the device thereof can simultaneously solve the camera movement and the time difference of each parallax image, so that the result after image fusion is clearer.

Embodiments of the present invention provide an image fusion method using multiple lenses, which is suitable for an image fusion device having multiple lenses. The image fusion method includes the following steps: simultaneously capturing a plurality of parallax images having different focal lengths through multiple lenses; analyzing a plurality of feature points of the plurality of parallax images; and calculating a matching relationship between the plurality of feature points of the plurality of parallax images; Each parallax image is shifted according to a matching relationship to adjust the same image portion of each parallax image to the same image position; and the result of displacing the plurality of parallax images is shifted, and a fused image is generated.

Embodiments of the present invention provide an image fusion device using a multi-lens, including a plurality of lenses, an image capture device, and an image processor. The image capture device electrically connects the plurality of lenses, and simultaneously captures a plurality of parallax images having different focal lengths through the plurality of lenses. The image processor is electrically connected to the image capturing device to receive the plurality of parallax images, and is configured to perform the following steps: analyzing a plurality of feature points of the plurality of parallax images; calculating a plurality of feature points of each parallax image and other parallax images a matching relationship of the plurality of feature points; shifting each of the parallax images according to the matching relationship to adjust the same image portion of each parallax image to the same image position; merging the results of the displacement of the plurality of parallax images, and generating a fused image And removing unwanted images from the edges of the fused image to produce a corrected fused image.

The detailed description of the present invention and the accompanying drawings are to be understood by the claims The scope is subject to any restrictions.

100‧‧‧Image fusion device

110‧‧‧ lens group

120‧‧‧Image capture device

130‧‧‧Image Processor

S210, S220, S230, S240, S250‧‧ steps

S232, S234, S236‧‧‧ steps

Ia, Ib‧‧ parallax images

D1a, D1b‧‧‧ image part

D2a, D2b‧‧‧ image part

S1, s2, s3, s4, s5, s6‧‧‧ feature points

R1, r2, r3, r4, r5, r6‧‧‧ feature points

P1‧‧‧Right direction

Wab‧‧‧Extra image

Iab‧‧‧ fusion image

Fab‧‧‧Fixed Fusion Image

S331, S332, S333, S334, S335, S336, S337‧‧ steps

Id, Ie, If‧‧ parallax images

D1d, D1e, D1f‧‧‧ image part

D2d, D2e, D2f, D3e‧‧‧ image parts

X1, x2, x3, x4, x5, x6‧‧‧ feature points

Y1, y2, y3, y4, y5, y6‧‧‧ feature points

Z1, z2, z3, z4, z5, z6‧‧‧ feature points

P2, P3‧‧‧ right direction

Wde, Wef‧‧‧ redundant images

Fde, Fef‧‧ image part

Ide, Ief‧‧ ‧ sub-image

M1, m2, m3, m4, m5, m6‧‧‧ sub-feature points

N1, n2, n3, n4, n5, n6‧‧‧ sub-feature points

Wdef‧‧‧Extra image

Idef‧‧‧ fusion image

Fdef‧‧‧corrected fusion image

1 is a schematic diagram of an image fusion device using a multi-lens according to an embodiment of the present invention.

2 is a flow of an image fusion method using a multi-lens according to an embodiment of the present invention; Cheng Tu.

3 is a detailed flow chart of shifting each parallax image in accordance with an embodiment of the present invention.

4A-4D are schematic diagrams of an image fusion method according to an embodiment of the present invention.

Figure 5 is a detailed flow chart of shifting each parallax image in accordance with another embodiment of the present invention.

6A-6E are schematic diagrams of an image fusion method according to another embodiment of the present invention.

In the following, the invention will be described in detail by way of illustration of various exemplary embodiments of the invention. However, the inventive concept may be embodied in many different forms and should not be construed as being limited to the illustrative embodiments set forth herein. In addition, the same reference numerals may be used in the drawings to indicate similar elements.

The image fusion method and device using the multi-lens according to the embodiment of the present invention are to simultaneously capture a plurality of parallax images having different focal lengths through multiple lenses, and according to the matching relationship between the feature points of each parallax image. (such as position matching relationship), adjust the same image portion of each parallax image to the same image position. Finally, the adjusted parallax image is blended to produce a fused image. Accordingly, the multi-lens image fusion method and the device thereof can simultaneously solve the camera movement and the time difference of each parallax image without artifact or blurring, so that the result after image fusion is clearer. In addition, since the multi-lens image fusion method and the device thereof simultaneously capture multiple parallax images (that is, there is no time difference problem for each parallax image), good image fusion can be obtained even if the moving object is photographed. result. The image fusion method using the multi-lens and the apparatus thereof disclosed in the present invention will be further described below.

First, please refer to FIG. 1, which shows a schematic diagram of an image fusion device using a multi-lens according to an embodiment of the present invention. As shown in FIG. 1 , the image fusion device 100 is configured to simultaneously capture multiple images of an external object to use the captured image as a source of image fusion. In this embodiment, the image fusion device 100 can be a smart phone and recorded. The present invention does not limit the video camera, the tablet computer, the notebook computer, or other devices that need to perform image fusion.

The image fusion device 100 includes a lens group 110, an image capture device 120, and an image processor 130. As shown in FIG. 1, the lens group 110 has a plurality of lenses (not shown in the drawings). In this embodiment, the plurality of lenses are disposed on the same plane (not shown in the drawings) and have a predetermined distance between the lenses.

The image capturing device 120 electrically connects the plurality of lenses of the lens group 110, and simultaneously captures the parallax images having different focal lengths through the plurality of lenses. For example, the lens group 110 has two lenses, and the image capturing device 120 simultaneously captures two images having different focal lengths as two parallax images through two lenses, such as the parallax images Ia and Ib shown in FIG. 4A. In each parallax image, the focal length corresponds to the sharpest image (such as the solid line). The farther away from the focal length, the more blurred the image (such as the dotted line). For example, in the parallax image Ia, the position of the focal length falls on the left human figure (ie, the solid line portion), and the farther away from the focal length, the more blurred the image will be (such as the right human figure in the dotted line portion); and in the parallax image Ib, the focal length The position falls on the right human form (ie, the solid line part), and the farther away from the focal length, the more blurred the image will be (such as the left human figure in the dotted line). And because there is a distance between adjacent lenses, there will be a distance difference between the parallax images Ia and Ib. Therefore, the parallax images Ia and Ib will correspondingly form the same image portions D1a and D1b (ie, have the same image content), and different image portions D2a and D2b (ie, have different image contents), as shown in FIG. 4A.

The image processor 130 is electrically connected to the image capture device 120 and receives the parallax image transmitted by the image capture device 120. The image processor 130 is configured to perform the following steps to adjust the same image portion of the received parallax image to the same image position, and perform image fusion on the adjusted parallax image.

Please refer to FIG. 1-2 at the same time. FIG. 2 is a flowchart showing an image fusion method using a multi-lens according to an embodiment of the present invention. First, the image processor 120 analyzes a plurality of feature points of the parallax image (step S210). In this embodiment, the image processor 120 is scale-invariant feature transform (SIFT). The algorithm searches for feature points of the parallax image, and the manner of searching for the feature points is known to those of ordinary skill in the art, and therefore will not be described again. The plurality of feature points of the parallax image may also be searched by other algorithms, such as a corner detector, a speeded up robust feature, etc., which are not limited by the present invention. Certainly, the plurality of feature points may also be all the pixels of the parallax image, or other information that can represent the parallax image, which is not limited by the present invention.

For example, referring to FIG. 4A and FIG. 4B simultaneously, the image capturing device 120 simultaneously captures the parallax images Ia and Ib having different focal lengths through the two lenses and transmits them to the image processor 130. The relationship between the parallax images Ia and Ib has been described in the above example, and therefore will not be described again. Therefore, in the parallax image Ia, the position of the focal length falls on the left human figure, and has the same image portion D1a and the different image portion D2a; and in the parallax image Ib, the position of the focal length falls on the right human figure and has the same image Part D1b and different image portion D2b. The image processor 130 will analyze the feature points of the parallax image Ia through the SIFT algorithm. It is assumed that the image processor 130 analyzes that the parallax image Ia has six feature points s1, s2, s3, s4, s5, and s6, and the parallax image Ib has six feature points r1, r2, r3, r4, r5, and r6, as shown in the figure. Shown in 4B.

Next, the image processor 130 calculates a matching relationship between the feature points of each of the parallax images and the feature points of the other parallax images (step S220). In this embodiment, the image processor 130 uses a Random Sample Consensus (RANSAC) algorithm to find the most suitable matching position of the feature points of each parallax image, and the search method of the feature points is a technology. The field is known to the general knowledge and will not be described again. The matching relationship of the feature points of each parallax image can also be calculated by other image matching algorithms, which is not limited by the present invention.

According to the above example, the image processor 130 finds that the feature point s1 of the parallax image Ia is matched to the feature point r2 of the parallax image Ib and the feature point s2 of the parallax image Ia is matched to the feature point r3 of the parallax image Ib and the parallax image through the RANSAC algorithm. The feature point s5 of Ia matches the feature point r4 of the parallax image Ib, and the parallax shadow The feature point s6 like Ia is matched to the feature point r5 of the parallax image Ib. The feature points s3 and s4 of the parallax image Ia and the feature points r1 and r6 of the parallax image Ib do not have a matching relationship.

After calculating the matching relationship of the feature points of each parallax image (ie, step S220), the image processor 130 shifts each parallax image according to the matching relationship to adjust the same image portion of each parallax image to the same image position. (Step S230). According to the above example, the parallax image Ia has the same image portion D1a and the different image portion D2a, and the parallax image Ib has the same image portion D1b and different image portion D2b. Therefore, the image processor 130 adjusts the same image portion D1a of the parallax image Ia and the same image portion D1b of the parallax image Ib to the same image position according to the matching relationship.

Furthermore, please refer to FIG. 3 at the same time, which shows a detailed flow chart of shifting each parallax image according to an embodiment of the present invention. First, the image processor 130 selects one of the parallax images as the main image, and the same image portion of the main image is located at the image position, that is, the image processor 130 uses the image position of the same image portion of the main image as the reference position. (Step S232). Then, the image processor 130 calculates the displacement amount of each of the unselected parallax images according to the matching relationship. The displacement amount is associated with the distance moved to the image position by the same image portion of the corresponding parallax image (step S234). Finally, the image processor 130 shifts the corresponding parallax image according to each displacement amount (step S236).

According to the above example, the image processor 130 uses the parallax image Ia as the main image, and sets the image position where the same image portion D1a of the parallax image Ia is located as the reference position. Then, the image processor 130 calculates the displacement amount of the unselected parallax image Ib according to the matching relationship, and the displacement amount represents that the same image portion D1b of the parallax image Ib is moved to the reference position (the same image portion D1a of the parallax image Ia is located) The distance of the image position). Finally, the image processor 130 shifts the parallax image Ib to the right direction P1 according to the amount of movement of the parallax image Ib to adjust the same image portion D1b of the parallax image Ib to the reference position (ie, the same image of the parallax image Ia). Part of the image location where D1a is located). Of course, the image processor 130 can also use the parallax image Ib as the main image, and repeat the process of steps S232-S236 to adjust the same image portion D1a of the parallax image Ia to the image of the same image portion D1b of the parallax image Ib. The position is not limited by the present invention.

After the same image portion of each parallax image is adjusted to the same image position (step S230), the image processor 130 merges the result of the displacement of the parallax image to generate a fused image (step S240). That is, in the process of image fusion, the image processor 130 fuses the clear image portion (ie, the position corresponding to the focal length) in the same image portion of each parallax image into a fused image (ie, has each parallax image) Part of the clear image). The image fusion technology in this step is known to those of ordinary skill in the art, and will not be described again.

In accordance with the above example, as shown in FIG. 4C, the image processor 130 shifts the result of merging the parallax images Ia and Ib to generate a fused image Iab. In the process of image fusion, the image processor 130 fuses the clear image portion of the same image portion D1a and D1b (ie, the left human figure in the same image portion D1a and the right human shape in the same image portion D1b) into one fusion. The image Iab is such that the fused image Iab has a clear image portion in each of the parallax images Ia and Ib, so that the fused image Iab is clearer than the parallax images Ia and Ib.

It should be noted that, since the image processor 130 is displaced according to the parallax image (as shown in FIG. 4B, the same image portion D1b of the parallax image Ib is displaced to the right direction P1 to the image of the same image portion D1a of the parallax image Ia. Position) produces a fused image, so the fused image forms an extra image left behind by the displacement of the parallax image. Therefore, after generating the fused image Iab (ie, step S240), the image processor 130 removes the extra image of the edge of the fused image to generate the corrected fused image for the image processor 130 to produce a better fused image. (Step S250). In accordance with the above example, as shown in FIGS. 4C and 4D, the image processor 130 will further remove the extra image Wab (in this embodiment, a blank image) of the edge of the fused image Iab to generate the corrected fused image Fab.

In addition, the image processor 130 may perform the method of shifting each parallax image according to the matching relationship to adjust the same image portion of each parallax image to the same image position, or may be performed in other manners, as another embodiment described below.

For convenience of description, the following example uses the image capture device 120 to simultaneously capture three images having different focal lengths as three parallax images through three lenses. As shown in the parallax images Id, Ie, and If shown in FIG. 6A, the image corresponding to the focal length is the clearest (such as the solid line portion), and the farther away from the focal length, the more blurred the image (such as the dotted line portion). For example, in the parallax image Id, the position of the focal length falls on the leftmost human shape (ie, the solid line portion), and the farther away from the focal length, the more blurred the image will be (such as the middle and right human figures of the dotted line); in the parallax image Ie The position of the focal length falls on the middle human figure (ie, the solid line part), and the farther away from the focal length, the more blurred the image will be (such as the left and right human figures of the dotted line); and in the parallax image If, the position of the focal length falls to the right. The human form (that is, the solid line part), and the farther away from the focal length, the more blurred the image will be (such as the left side of the dotted line and the middle human figure). In addition, there is a distance difference between the parallax images Id, Ie, and If due to a distance between adjacent lenses. Therefore, the parallax images Id and Ie will correspondingly form the same image portions D1d, D1e, and D3e (i.e., have the same image content) and different image portions D2d and D2e (i.e., have different image contents). The parallax images Ie and If will correspondingly form the same image portions D1e, D2e, and D1f (i.e., have the same image content) and different image portions D3e and D2f (i.e., have different image contents).

Next, as shown in FIG. 6B, the image processor 130 analyzes the parallax image Id by the SIFT algorithm to have six feature points x1, x2, x3, x4, x5, and x6, and the parallax image Ie has six feature points y1 and y2. , y3, y4, y5, and y6, and the parallax image If have six feature points z1, z2, z3, z4, z5, and z6 (as in the above-described step S210). Then, the image processor 130 matches the feature points x1-x6 of the parallax image Id by the RANSAC algorithm to the feature points y1-y6 of the parallax image Ie, and the feature points x1-x6 of the parallax image Id are respectively matched to the parallax image. The feature points z1-z6 of If (like step S220 above).

After step S220, the image processor 130 shifts each parallax image in a manner different from that of steps S232-S234 of FIG. 3 to adjust the same image portion of each parallax image to the same image position. as follows.

Please refer to FIG. 5, which shows a detailed flow chart of shifting each parallax image according to another embodiment of the present invention. First, the image processor 130 selects two of the plurality of parallax images as the image group to be matched, and each of the parallax images is selected at least once (step S331). For example, as shown in FIG. 6B, the image processor 130 selects the parallax images Id and Ie as the first group of to-be-matched image groups, and selects the parallax images Ie and Id as the second group of to-be-matched image groups.

Next, the image processor 130 selects one of the parallax images as the first image in each of the to-be-matched image groups, and selects one of the other parallax images as the second image. The first image and the second image have the same sub-image (step S332). According to the above example, as shown in FIG. 6B, in the first group of to-be-matched image groups, the parallax image Id is selected as the first image, the parallax image Ie is used as the second image, and the parallax image Id and the parallax image Ie have the same sub-image. That is, the same image portions D1d, D1e, and D3e. The parallax image Ie is selected as the first image, and the parallax image If is the second image, and the parallax image If has the same sub-image, that is, the same image portion D1e, D2e and D1f.

Then, the image processor 130 calculates a first displacement amount of the second image according to the matching relationship in each image group to be matched. The first displacement amount is related to the distance that the sub-image of the second image is moved to the sub-image of the first image (step S333). According to the above example, as shown in FIG. 6B, in the first group of to-be-matched image groups, the image processor 130 calculates a first displacement amount of the second image (ie, the parallax image Ie) according to the matching relationship, and the first displacement amount represents The sub-images of the second image (ie, the same image portions D1e and D3e of the parallax image Ie) are moved to the distance of the sub-images of the first image (ie, the same image portion D1d of the parallax image Id). In the second group of to-be-matched image groups, the image processor 130 calculates a first displacement amount of the second image (ie, the parallax image If) according to the matching relationship, and the first displacement amount represents the sub-image of the second image (ie, the parallax image If Phase The same image portion D1f) is moved to the distance of the sub-image of the first image (i.e., the image portions D1e and D3e having the same parallax image Ie).

Next, the image processor 130 shifts the second image according to the first displacement amount in each image group to be matched (step S334). According to the above example, as shown in FIG. 6B, in the first group of to-be-matched image groups, the image processor 130 shifts the parallax image Ie to the right direction P2 according to the first displacement amount of the parallax image Ie to make the parallax image Ie the same. The image portions D1e and D3e are adjusted to the image position where the same image portion D1d of the parallax image Id is located. In the second group of images to be matched, the image processor 130 shifts the parallax image If to the right direction P2 according to the first displacement amount of the parallax image If to adjust the same image portion D1f of the parallax image If to the parallax image Ie. The image position where the same image portions D1e and D2e are located.

Next, the image processor 130 merges the results of the displacement of the image group to be matched in each image group to be matched to respectively become a sub-fused image, wherein each sub-fused image has the same image portion (step S335). According to the above example, as shown in FIG. 6C, the image processor 130 generates the sub-fused image Ide according to the result of the displacement of the fused parallax images Id and Ie, and combines the results of the disparity images Ie and If to generate the sub-sequence. Fusion image Ief. At this time, the sub-fused image Ide and the sub-fused image Ief will have the same image portions Fde and Fef. The sub-fused images Ide and Ief also have redundant portions Wde and Wef (in this embodiment, blank images).

Then, the image processor 130 analyzes the sub-feature points of each sub-fused image, and calculates a sub-match relationship between the sub-feature points of each sub-fused image (step S336). According to the above example, as shown in FIG. 6C, the image processor 130 analyzes the sub-fused image Ide through the SIFT algorithm to have six sub-feature points m1, m2, m3, m4, m5, and m6, and the sub-fused image Ief has six sub-segments. Feature points n1, n2, n3, n4, n5 and n6. The image processor 130 then finds the sub-feature points m1-m6 of the sub-parallax image Ide through the RANSAC algorithm to match the sub-feature points n1-n6 of the parallax image Ief, that is, the sub-matching relationship.

Finally, the image processor 130 shifts each sub-fused image according to the sub-matching relationship to adjust the same image portion of each sub-fused image to the same image position (step S337). In accordance with the above example, the sub-fused images Ide and Ief have the same image portions Fde and Fef. Therefore, the image processor 130 adjusts the same image portion Fde of the sub-fused image Ide and the same image portion Fef of the sub-fused image Ief to the same image position according to the sub-matching relationship.

The step of the image processor 130 shifting each sub-fused image according to the sub-matching relationship can be roughly derived from steps S232-S236, FIG. 3 and FIG. 4B. First, the image processor 130 selects one of the sub-fused images as a main image, and the same image portion of the main image is located at the image position, that is, the image processor 130 uses the image position of the same image portion of the main image as a reference. Positioning; further, the image processor 130 calculates a second displacement amount of each of the sub-fusion images that are not selected according to the sub-matching relationship, and the second displacement amount is associated with the same image portion of the corresponding sub-fused image to move to the image position. Finally, the image processor 130 will shift the corresponding parallax image according to each displacement amount.

According to the above example, the image processor 130 sets the sub-fused image Ide as the main image, and sets the image position where the same image portion Fde of the sub-fused image Ide is located as the reference position. Then, the image processor 130 calculates a second displacement amount of the unselected sub-fused image Ief according to the sub-matching relationship, and the second displacement amount represents the same image portion Fef of the sub-fused image Ief moves to the reference position (sub-fused image) The distance of the image position where the same image portion of the Ide is located. Finally, the image processor 130 shifts the sub-fused image Ief to the right direction P3 according to the amount of movement of the sub-fused image Ief to adjust the same image portion Fef of the sub-fused image Ief to the reference position (ie, the same image portion of the sub-fused image Ide) The image location where Fde is located).

After the same image portion of each sub-fused image is adjusted to the same image position (ie, step S337), the image processor 130 performs step S240 to fuse the results of the displacement of the plurality of parallax images, and thereby generate the fusion. image. In accordance with the above example, as shown in FIG. 6D, the image processor 130 will fuse the disparity image Id, The result of the displacement of Ie and If (the result of the sub-fusion image Ide and Ief displacement) is used to generate the fused image Idef.

After the image processor 130 generates the fused image, the image processor 130 removes the redundant image of the edge of the fused image to generate the corrected fused image for the image processor 130 to generate a better fused image. According to the above example, as shown in FIGS. 6D and 6E, the image processor 130 further removes the extra image Wdef (two blank images in this embodiment) of the edge of the fused image Idef to generate the corrected fused image Fdef. .

In summary, the image fusion method and device using the multi-lens according to the embodiments of the present invention are capable of simultaneously capturing a plurality of parallax images having different focal lengths through multiple lenses, and the same image of each parallax image. Partially adjusted to the same image position. Finally, the adjusted parallax image is blended to generate a fused image. Accordingly, the multi-lens image fusion method and the device thereof can simultaneously solve the camera movement and the time difference of each parallax image without artifact or blurring, so that the result after image fusion is clearer. In addition, since the multi-lens image fusion method and the device thereof simultaneously capture multiple parallax images (that is, there is no time difference problem for each parallax image), good image fusion can be obtained even if the moving object is photographed. result.

The above description is only an embodiment of the present invention, and is not intended to limit the scope of the invention.

S210, S220, S230, S240, S250‧‧ steps

Claims (10)

An image fusion method using a multi-lens is applicable to an image fusion device having a plurality of lenses. The image fusion method includes: capturing a plurality of parallax images having different focal lengths through the lenses; and analyzing the plurality of parallax images a feature point; calculating a matching relationship between the feature points of the parallax images; and shifting each of the parallax images according to the matching relationship to adjust one of the same image portions of each of the parallax images to the same image Positioning the result of the displacement of the parallax images and generating a fused image; wherein, in the step of displacing each of the parallax images, the method further comprises: selecting one of the parallax images as a main image, and the main image The same image portion of the image is located at the image position; and a displacement amount of each of the unselected parallax images is calculated according to the matching relationship, wherein the displacement amount is associated with the corresponding image portion of the corresponding parallax image to move to the image position The distance; and the parallax image corresponding to each displacement amount. The image fusion method of claim 1, wherein the lenses are disposed on a same plane and have a predetermined distance between adjacent ones of the lenses. The image fusion method of claim 1, wherein after the generating the fused image, the method further comprises: removing an extra image of the edge of the fused image to generate the fused image. An image fusion method using a multi-lens is applicable to an image fusion device having a plurality of lenses. The image fusion method includes: capturing a plurality of parallax images having different focal lengths through the lenses; and analyzing the plurality of parallax images Feature points; Calculating a matching relationship between the feature points of the parallax images; and shifting each of the parallax images according to the matching relationship to adjust one of the same image portions of each of the parallax images to the same image position; The result of the displacement of the parallax images and the generation of a fused image; wherein, in the step of displacing each of the parallax images, the method further comprises: respectively selecting two of the parallax images as a to-be-matched image group, and each of the images The parallax image is selected at least once; in each of the to-be-matched image groups, one of the parallax images is selected as a first image, and the other of the parallax images is selected as a second image, and the first image and the first image The second image has the same sub-image; in each of the image groups to be matched, a first displacement amount of the second image is calculated according to the matching relationship, wherein the first displacement amount is associated with the sub-image of the second image Moving the image to the distance of the sub-image of the first image; in each of the to-be-matched image groups, shifting the second image according to the first displacement amount; In the matched image group, the result of the displacement of the image group to be matched is merged to become a sub-fused image, wherein each of the sub-fused images has the same image portion; and a plurality of sub-feature points of the sub-fused images are analyzed, and Calculating a sub-matching relationship between the sub-feature points of the sub-fused images; and shifting each of the sub-fused images according to the sub-matching relationship to adjust the same image portion of each of the sub-fused images to the image position. The image fusion method of claim 4, wherein the step of displacing each of the sub-fused images further comprises: selecting one of the sub-fused images as a primary image, and the same image portion of the primary image is located The image position; calculating, according to the sub-matching relationship, one of the unselected ones of the sub-fused images a displacement amount, wherein the second displacement amount is associated with a distance at which the same image portion of the corresponding sub-fused image moves to the image position; and the corresponding sub-fused image is displaced according to each of the second displacement amounts. The image fusion method of claim 4, wherein the lenses are disposed on a same plane and have a predetermined distance between adjacent ones of the lenses. The image fusion method of claim 4, wherein after the generating the fused image, the method further comprises: removing an extra image of the edge of the fused image to generate the fused image. An image fusion device using a multi-lens, comprising: a plurality of lenses; an image capture device electrically connecting the lenses, and simultaneously capturing a plurality of parallax images having different focal lengths through the lenses; an image processor, electricity Connecting the image capture device to receive the parallax images, and performing the following steps: analyzing a plurality of feature points of the parallax images; calculating the feature points of each of the parallax images and the other of the parallax images a matching relationship of the feature points; shifting each of the parallax images according to the matching relationship to adjust the same image portion of each of the parallax images to the same image position; and merging the results of the parallax image displacements, and generating a fused image; and a superimposed image of the edge of the fused image to be generated to generate the fused image; wherein the image processor performs displacement in each of the parallax images, and further includes: Selecting one of the parallax images as a main image, and the same image portion of the main image is located at the image position; and calculating a displacement amount of each of the unselected parallax images according to the matching relationship, where the displacement amount is associated And shifting the corresponding image portion of the corresponding parallax image to the image position; and shifting the corresponding parallax image according to each of the displacement amounts. An image fusion device using a multi-lens, comprising: a plurality of lenses; an image capture device electrically connecting the lenses, and simultaneously capturing a plurality of parallax images having different focal lengths through the lenses; an image processor, electricity Connecting the image capture device to receive the parallax images, and performing the following steps: analyzing a plurality of feature points of the parallax images; calculating the feature points of each of the parallax images and the other of the parallax images a matching relationship of the feature points; shifting each of the parallax images according to the matching relationship to adjust the same image portion of each of the parallax images to the same image position; and merging the results of the parallax image displacements, and generating a fused image; and a superimposed image of the edge of the fused image to be generated to generate the fused image; wherein the image processor performs displacement of each of the parallax images, further comprising: respectively selecting the images Two of the parallax images are used as a to-be-matched image group, and each of the parallax images is selected at least once; in each of the to-be-matched image groups Wherein the selecting a parallax image as a first image, wherein the other of the selected image as a second parallax Movies And the first image and the second image have the same sub-image; in each of the to-be-matched image groups, calculating a first displacement amount of the second image according to the matching relationship, wherein the first displacement The amount of the sub-image associated with the second image is moved to the distance of the sub-image of the first image; in each of the to-be-matched image groups, the corresponding second image is displaced according to the first displacement amount; In the image group to be matched, the result of the displacement of the image group to be matched is merged to become a sub-fused image, wherein each of the sub-fused images has the same image portion; and a plurality of sub-features of the sub-fused images are analyzed Pointing, and calculating a sub-matching relationship between the sub-feature points of the sub-fused images; and shifting each of the sub-fused images according to the sub-matching relationship to adjust the same image portion of each of the sub-fused images To the image location. The image fusion device of claim 9, further comprising: performing, by the image processor, the displacement of each of the sub-fused images, further comprising: selecting one of the sub-fused images as a primary image, and the primary image The same image portion is located at the image position; and a second displacement amount of each of the un-selected sub-fused images is calculated according to the sub-matching relationship, wherein the second displacement amount is associated with the corresponding image of the corresponding sub-fused image a portion of the distance moved to the image position; and the corresponding sub-fused image is shifted according to each of the second displacement amounts.