CN102779329B - Image processing apparatus and image processing method - Google Patents

Abstract

The present invention proposes an image processing device and an image processing method. The image processing device includes an image correction module, an object motion detection module and an image mixing module. The image correction module estimates the block displacement and global displacement of the unselected image relative to the target image, and performs displacement correction to generate a corrected image. The object motion detection module determines whether the difference between the displacement of each block and the global displacement is greater than a threshold value, thereby generating an object motion pointer. The image mixing module performs arithmetic operations on each pixel of the target image and each pixel of the corrected image according to the object motion pointer to generate a super-resolution image. The present invention can enlarge a single or multiple low-resolution images to generate a high-resolution image.

Description

Image processing device and image processing method

technical field

The present invention relates to an image processing device and an image processing method thereof, and in particular to an image processing device and an image processing method for super-resolution (super-resolution) image enlargement.

Background technique

Image magnification technology is an important research direction in image processing, and image interpolation is a method related to image magnification technology. Generally speaking, image interpolation is only completed by enlarging a single image. Common single image enlarging methods include polynomial interpolation, edge-directed interpolation, and sample-based super-analysis. Technology (exampled-basedforsuper-resolution).

Although the polynomial interpolation method is simple and has a fast operation speed, it often causes the enlarged image to be blurred due to the lack of high-frequency information of the image, and block effect will occur. Edge-wise interpolation or sample-based super-resolution techniques require a huge amount of computation. Therefore, the image quality is greatly limited in the technique of image enlargement from a single image enlargement.

However, if you want to use the technology of mixing multiple image sequences to produce enlarged images, the most common side effect is ghosting. The cause of the ghosting phenomenon is that when taking continuous shots of the same scene, individual objects in the scene are moving, and when mixing images, the overall displacement (or camera displacement) of the image is corrected, which does not affect the overall displacement of the scene. Therefore, if there are individual moving objects in the image, it will cause ghosting in the blended image.

Contents of the invention

In view of this, the present invention provides an image processing device that can enlarge multiple low-resolution images to generate high-resolution images, and first detect whether there is movement of individual objects in the images, and then mix multiple corrected images to output super-resolution images rate image.

The present invention also provides an image processing method, which is used to process multiple low-resolution images and enlarge them to generate high-resolution images, and can mix multiple corrected images to generate super-resolution images.

An image processing device proposed by the present invention is used to receive a plurality of second-resolution images generated by enlarging a plurality of first-resolution images, wherein one of the second-resolution images is a target image, and the remaining The second resolution image is a plurality of unselected images. The image processing device includes an image correction module, an object movement detection module and an image mixing module. Wherein, the image correction module estimates multiple block displacements and multiple global displacements of the unselected image relative to the target image, and performs multiple displacement corrections on the unselected image, so as to generate multiple corrected images. The object movement detection module is coupled to the image correction module to determine whether the difference between the displacement of each block and the corresponding global displacement is greater than a threshold value, so as to generate a plurality of object movement pointers. The image mixing module is coupled to the object movement detection module, and the image mixing module performs an arithmetic operation on each pixel of the target image and each pixel of the corrected image according to the object movement pointer to generate a third resolution image. The resolution of the high-resolution image is higher than the resolution of the second-resolution image.

In an embodiment of the present invention, the image blending module sets a plurality of weights according to the moving pointers of the object, and the arithmetic operation is that the image blending module uses the weights to combine each pixel of the target image with the Each pixel of these corrected images is weighted and summed.

In an embodiment of the present invention, the image mixing module performs multiple directional gradient operations on each pixel of the target image and each pixel of the corrected image to generate multiple gradient differences.

In one embodiment of the present invention, when the difference between the displacement of each block and the corresponding global displacement is greater than the threshold value, the object movement detection module drives the object movement pointer, and when the displacement of each block is greater than the corresponding When the difference in global displacement is smaller than the threshold value, the object movement detection module blocks the object movement pointer.

In an embodiment of the present invention, when the object moving pointer is in the driving state, the image mixing module sets the weighting to zero, and when the object moving pointer is in the blocked state, the image mixing module sets the weighting to gradient difference.

In an embodiment of the present invention, the image mixing module performs a direction gradient operation on each pixel of the target image and each pixel of the corrected image to generate a plurality of direction gradient values, and the direction gradient values include horizontal Directional gradient values, vertical gradient values, and diagonal gradient values.

In an embodiment of the present invention, the image mixing module selects the direction gradient value with the largest numerical value as the maximum gradient value for each pixel point of the target image and each pixel point of the corrected image, and selects the direction The smallest gradient value is used as the minimum gradient value, and each gradient difference is equal to the difference between the maximum gradient value and the minimum gradient value.

In an embodiment of the present invention, the image correction module includes a block displacement estimation unit and a global displacement estimation unit. The block displacement estimating unit divides the target image and the unselected image into a plurality of blocks, and estimates the block displacement of the unselected image relative to the target image. The global displacement estimating unit is coupled to the block displacement estimating unit, and executes a plurality of global displacement estimations according to the block displacement, so as to generate the global displacement.

In an embodiment of the present invention, the image correction module includes a displacement correction unit, which uses an affine transformation matrix (Affine transformation matrix) to perform displacement correction, so that the starting point position of the unselected image is corrected to the starting point of the target image same location.

From another point of view, an image processing method proposed by the present invention is used to process a plurality of second-resolution images generated by enlarging a plurality of first-resolution images, wherein one of the second-resolution images One is a target image, and the remaining second-resolution images are multiple unselected images. The image processing method includes the following steps: estimating multiple block displacements and multiple global displacements of the unselected image relative to the target image, performing multiple displacement corrections on the unselected image, so as to generate multiple corrected images. In addition, it is judged whether the difference between the displacement of each block and the corresponding global displacement is greater than a threshold value, so as to generate a plurality of object movement pointers. Furthermore, an arithmetic operation is performed on each pixel of the target image and each pixel of the corrected image according to the moving pointer of the object to generate a third resolution image, and the resolution of the third resolution image is higher than the second resolution The resolution of the image.

Based on the above, the present invention can enlarge a single or multiple low-resolution images to generate a high-resolution image, and use multiple images to mix to generate a super-resolution image with high quality and rich detail information. Before image blending, it will be judged whether there are individual objects in the image moving, so as to avoid ghosting in the blended super-resolution image.

In order to make the above-mentioned features and advantages of the present invention more comprehensible, the following specific embodiments are described in detail with reference to the accompanying drawings.

Description of drawings

FIG. 1 is a block diagram of an image processing device according to an embodiment of the present invention.

FIG. 2 is a block diagram of an image processing device according to another embodiment of the present invention.

FIG. 3 is a flowchart of a method for performing object movement detection by the object movement detection module 200 according to an embodiment of the present invention.

FIG. 4 is a flowchart of a method for performing image blending by the image blending module 300 for one of the pixels according to an embodiment of the present invention.

FIG. 5 is a flowchart of an image processing method according to an embodiment of the present invention.

Reference signs:

10: Image processing device;

100: image correction module;

110: block displacement estimation unit;

120: global displacement estimation unit;

130: displacement correction unit;

200: object movement detection module;

300: image mixing module;

Img1_LR～Img4_LR: low-resolution images;

Img1_HR～Img4_HR: high resolution image;

CorrImg2_HR～CorrImg4_HR: corrected image;

Img1_SR: super-resolution image;

Map2～Map4: object moving pointer;

S310～S370: steps of the object movement detection method;

S410-S470: the step of performing image mixing on each pixel;

S510-S530: Steps of the image processing method.

detailed description

Please refer to FIG. 1 first. FIG. 1 is a block diagram of an image processing device according to an embodiment of the present invention. The image processing device 10 is used to receive a plurality of second-resolution images generated by enlarging a plurality of first-resolution images, wherein the plurality of first-resolution images can be obtained by a digital image capture device of a CMOS sensor such as a digital Captured by cameras, digital video cameras (Digital Video, DV), etc. The CMOS sensor is characterized by high-speed continuous shooting, so it can continuously capture multiple first-resolution images of a scene. In addition, those skilled in the art may use well-known image enlargement methods to enlarge multiple first-resolution images to generate multiple second-resolution images according to actual requirements. Image enlargement methods include using polynomial interpolation methods, interpolation methods along the edge direction, or sample-based super-analysis methods and so on. Here, one of the second-resolution images must be selected as the target image, and the remaining second-resolution images are a plurality of unselected images.

The image processing device 10 includes an image correction module 100 , an object movement detection module 200 and an image mixing module 300 . A plurality of first-resolution images are taken continuously for a scene. When hand-held image processing device 10 shoots and hand shake occurs, there will be sub-pixel shift (sub-pixel shift) between the first-resolution images. The image correction module 100 can respectively estimate the multiple block displacements of each unselected image relative to the target image, and use these block displacements to determine the global displacement of each unselected image relative to the target image, and the image correction module 100 performs displacement correction on the unselected images according to the global displacement, so as to generate a plurality of corrected images.

The object movement detection module 200 is coupled to the image correction module 100 to determine whether the difference between the displacement of each block and the corresponding global displacement is greater than a threshold value, so as to generate a plurality of object movement pointers. The threshold value can be pre-determined by the user according to actual needs, and no limitation is imposed here. Next, the image mixing module 300 coupled to the object movement detection module 200 can perform an arithmetic operation on each pixel of the target image and each pixel of the corrected image according to the object movement pointer to generate a third resolution image. The resolution of the triple resolution image is higher than the resolution of the second resolution image.

That is to say, when the difference between the displacement of each block and the corresponding global displacement is greater than the threshold value, the object movement detection module 200 drives the generated object movement pointer. And this driven object movement pointer represents that there is a phenomenon of object movement in this block. The object movement detection module 200 instructs the image blending module 300 not to perform image blending for this block through the driven object movement pointer. On the contrary, when the difference between the displacement of each block and the corresponding global displacement is not greater than the threshold value, the object movement detection module 200 blocks the generated object movement pointer to represent that there is no object movement in this block , so can participate in the action of image blending. Accordingly, the problem of image ghosting as described in the prior art can be effectively solved.

In order to further illustrate the detailed operation of the image processing device 10 and to make the content of the present invention more clear, another embodiment is exemplified below as an example in which the present invention can indeed be implemented.

FIG. 2 is a block diagram of an image processing device according to another embodiment of the present invention, please refer to FIG. 2 . The image correction module 100 of the image processing device 10 includes a block displacement estimation unit 110 , a global displacement estimation unit 120 and a displacement correction unit 130 . Firstly, the block displacement estimating unit 110 is used to receive 4 second resolutions ( This embodiment is exemplified by high-resolution) images Img1_HR, Img2_HR, Img3_HR and Img4_HR. In this embodiment, four received images are taken as an example for illustration, but the present invention does not limit the number of received images.

The block displacement estimation unit 110 divides the target image Img1_HR and the unselected images Img2_HR, Img3_HR, and Img4_HR into multiple blocks in the same way. For example, if the size of the target image and the unselected images is P×Q, then Divide the target image and unselected images into M×N blocks, where M, N, P, and Q are integers greater than 1, and M is less than or equal to P, and N is less than or equal to Q. The division method can be set according to actual needs and is not limited. Next, the block displacement estimating unit 110 estimates the block displacement of each block of the unselected images Img2_HR, Img3_HR, and Img4_HR relative to the target image Img1_HR. The estimation method is, for example, block comparison.

The global displacement estimation unit 120 respectively performs global displacement estimation on each of the unselected images Img2_HR, Img3_HR and Img4_HR. For example, the method for estimating the global displacement can take the mode of the displacements of multiple blocks, that is, the block displacements are counted first, and the block displacement that occurs the most times is selected as the global displacement, or All block displacements are averaged to obtain the global displacement and so on. Therefore, each of the non-selected images Img2_HR, Img3_HR and Img4_HR has its own global displacement.

The displacement correction unit 130 performs displacement correction according to the above-mentioned global displacement. The displacement correction uses an affine transformation matrix (Affine transformation matrix) to correct the starting point positions of the unselected images Img2_HR, Img3_HR, and Img4_HR to the starting point positions of the target image Img1_HR. same starting point. The affine transformation matrix can be used for rotation and movement correction, and the coefficients of the matrix can be obtained from the global displacement. Based on the above, the corrected images CorrImg2_HR, CorrImg3_HR and CorrImg4_HR can be obtained.

The object movement detection module 200 performs object movement detection on each of the corrected images CorrImg2_HR, CorrImg3_HR and CorrImg4_HR and the target image Img1_HR to generate object movement pointers Map2, Map3 and Map4. FIG. 3 is a flowchart of a method for performing object movement detection by the object movement detection module 200 according to an embodiment of the present invention. Please refer to FIG. 2 and FIG. 3 together. As shown in step S310, object movement detection is performed from the first pixel of the image, since the block displacement and the global displacement can be obtained in the block displacement estimation unit 110 and the global displacement estimation unit 120, therefore In step S320, the difference Diffi between the block displacement and the global displacement can be calculated, and the calculation formula is as follows:

Diff _i ＝|X_LM _i -X_GM|+|Y_LM _i -Y_GM|

Among them, X_LM _i and Y_LM _i respectively represent the horizontal component and vertical component of the block displacement, i represents the i-th block of the image to which this pixel belongs, i is a positive integer greater than zero, X_GM, Y_GM represent the global displacement of the image The horizontal and vertical components of the quantity.

Next, in step S330, it is judged whether the difference Diff _i is greater than a threshold TH, which can be preset by those skilled in the art according to the actual situation. If the difference Diff _i is greater than the threshold value TH, then continue to step S340, the object movement detection module 200 drives the object movement pointer (for example, setting the object movement pointer to 1) to represent that there is object movement in this block. If not, the object movement detection module 200 blocks the object movement pointer (for example, setting the object movement pointer to 0), which means that there is no object movement in this block. Each pixel in the image needs to be judged through this process, so step S360 judges whether it is the last pixel in the image, if so, end the object movement detection, if not, enter the calculation and judgment of the next pixel.

The image mixing module 300 mixes each pixel in the target image Img1_HR and the corrected images CorrImg2_HR, CorrImg3_HR, and CorrImg4_HR. During the mixing process, the pointers Map2, Map3, and Map4 must be moved in conjunction with the reference object to generate the third resolution (this embodiment An example is the super-resolution) image Img1_SR. For example, the first pixel of the super-resolution image Img1_SR is obtained by mixing the first pixel of the target image Img1_HR with the first pixels of the corrected images CorrImg2_HR, CorrImg3_HR, and CorrImg4_HR.

For the detailed mixing method, please refer to FIG. 2 and FIG. 4 at the same time. FIG. 4 is a flowchart of a method for performing image mixing on a single pixel by the image mixing module 300 according to an embodiment of the present invention.

As shown in step S410, starting from the first image (for example, the target image Img1_HR), the image mixing module 300 performs a direction gradient operation on this pixel to generate a direction gradient value, which includes a horizontal gradient value H_Gra, a vertical gradient value value V_Gra and two diagonal gradient values D−_Gra, D+_Gra (step S420). Wherein, the horizontal direction gradient value H_Gra is the sum of the absolute values of grayscale differences between this pixel point and two adjacent horizontal direction pixel points. The vertical gradient value V_Gra is the sum of the absolute values of the grayscale differences between this pixel and two adjacent vertical pixels. The diagonal direction gradient values D-_Gra, D+_Gra include the sum of the absolute value of the gray scale difference between this pixel point and two adjacent first diagonal pixel points and the gray scale difference between this pixel point and two adjacent second diagonal The sum of the absolute values of the gray scale differences of pixels in the direction.

Then in step S430 , the image blending module 300 further selects the largest value among the above-mentioned direction gradient values as the maximum gradient value Max_Gra, and selects the smallest value among the above-mentioned direction gradient values as the minimum gradient value Min_Gra. Step S440 calculates the gradient difference Diff_Gra between the maximum gradient value Max_Gra and the minimum gradient value Min_Gra.

Step S450 judges whether it is the last image, if not, enters the next image, and does not enter step S470 until the gradient difference Diff_Gra of each image to be mixed is calculated. In order to avoid the generation of ghost images, the image mixing module 300 must refer to the object movement pointers Map2, Map3 and Map4. When the object movement pointers Map2, Map3 and Map4 are 1 at this pixel, it means that there is an object moving, and then the weighted Weight If it is set to 0, the pixel will not be blended. On the contrary, when the object movement pointers Map2, Map3 and Map4 are set to 0 at the pixel, the image mixing module 300 will set the weighting to the gradient difference Diff_Gra of this pixel, and the larger the gradient difference Diff_Gra, the image has texture Or the existence of an edge indicates that the image information of this pixel is more important and needs to be preserved, so the gradient difference Diff_Gra is used as the weight. The image mixing module 300 utilizes the above-mentioned weighting to carry out a weighted sum to the pixels of each image, wherein the calculation formula of the weighted sum FV is as follows:

$\mathrm{<span\; class="notranslate">\; FV</span>}<span\; class="notranslate">\; =</span>\frac{\underset{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; Weight</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; img</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; )</span>}{\underset{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; Weight</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ]</span>}$

Among them, n represents the nth image, Weight[n] represents the weight of one of the pixels in the nth image, and img[n] represents the grayscale value of one of the pixels in the nth image. Therefore, the super-resolution image Img1_SR can be output after the target image Img1_HR and each pixel in the corrected images CorrImg2_HR, CorrImg3_HR, and CorrImg4_HR are mixed as described above.

From another point of view, FIG. 5 is a flow chart of an image processing method according to an embodiment of the present invention, which is used to process multiple second-resolution images generated by enlarging multiple first-resolution images , wherein one of the second-resolution images is the target image, and the remaining second-resolution images are a plurality of unselected images. Please refer to FIG. 5, as shown in step S510, estimate multiple block displacements and multiple global displacements of the unselected image relative to the target image, and perform multiple displacement corrections on the unselected image, thereby generating multiple corrections after image. In step S520, it is determined whether the difference between the displacement of each block and the corresponding global displacement is greater than a threshold value, so as to generate a plurality of object movement pointers. Finally, as described in step S530, an arithmetic operation is performed on each pixel of the target image and each pixel of the corrected image according to the moving pointer of the object to generate a third resolution image, and the resolution of the third resolution image is higher than The resolution of the second resolution image.

In summary, the present invention can magnify a single or multiple low-resolution images to produce a high-resolution image, and use multiple high-resolution images to mix to produce a super image with high quality and rich details. resolution image. Before doing image mixing, it will also detect whether there are individual objects moving in the image. In the image area where individual objects move, select a single image to enlarge, and do not perform image mixing actions, so as to avoid ghosts that are easy to occur when multiple images are mixed. shadow problem. For other areas where image blending is required, the gradient difference is used as weighting, so that the blended super-resolution image can maintain the sharpness of the image, and can simultaneously remove the block effect and achieve the effect of reducing noise.

Although the present invention has been disclosed above with the embodiments, it is not intended to limit the present invention, and any person skilled in the art may make some changes and modifications without departing from the spirit and scope of the present invention.

Claims (14)

1. An image processing device, used to receive a plurality of second resolution images generated by enlarging a plurality of first resolution images, wherein one of the second resolution images is a target image, and the remaining The second resolution images for are multiple unselected images, including: An image correction module, which estimates a plurality of block displacements and a plurality of global displacements of the unselected images relative to the target image, and performs multiple displacement corrections on the unselected images, thereby generating multiple corrected image; An object movement detection module, coupled to the image correction module, judges whether the difference between the displacement of each block and the corresponding global displacement is greater than a threshold value, so as to generate a plurality of object movement pointers, each of which The object moving pointer is one of a driving state and a blocking state. The driving state is that there is object movement in the block corresponding to the object moving pointer, and the blocking state is that the block corresponding to the object moving pointer there is no movement of objects; and An image mixing module, coupled to the object movement detection module, the image mixing module sets a plurality of weights according to the object movement pointers, for each pixel of the target image and each pixel of the corrected images performing a plurality of directional gradient operations to generate a plurality of gradient differences, and using the weights to each pixel of the target image and each pixel of the corrected images according to the object movement pointers and the gradient differences performing an arithmetic operation on the pixels to generate a third resolution image, wherein the resolution of the third resolution image is higher than the resolution of the second resolution images, wherein when the objects move the pointer is the driving state , then the image mixing module sets these weights to zero, and when the object moving pointers are in the blocked state, then the image mixing module sets these weights as the gradient difference values. 2. The image processing device according to claim 1, wherein the arithmetic operation is a weighted sum of each pixel of the target image and each pixel of the corrected images by the image mixing module using the weights. 3. The image processing device according to claim 2, wherein when the difference between each block displacement and the corresponding global displacement is greater than the threshold value, the object movement detection module drives the objects to move Pointer, when the difference between the displacement of each block and the corresponding global displacement is less than the threshold value, the object movement detection module blocks the object movement pointers. 4. The image processing device according to claim 2, wherein the image mixing module performs the direction gradient calculations for each pixel of the target image and each pixel of the corrected images to generate a plurality of directions Gradient values, the gradient values in the direction include a gradient value in the horizontal direction, a gradient value in the vertical direction and two gradient values in the diagonal direction. 5. The image processing device according to claim 4, wherein the image mixing module selects the one with the largest value among the direction gradient values for each pixel of the target image and each pixel of the corrected images As a maximum gradient value, and selecting the smallest value among the direction gradient values as a minimum gradient value, each gradient difference is equal to the difference between the maximum gradient value and the minimum gradient value. 6. The image processing device according to claim 1, wherein the image correction module comprises: a block displacement estimation unit, which divides the target image and the unselected images into a plurality of blocks, and estimates the displacement of the blocks of the unselected images relative to the target image; and A global displacement estimation unit, coupled to the block displacement estimation unit, executes a plurality of global displacement estimations according to the block displacements, so as to generate the global displacements. 7. The image processing device according to claim 1, wherein the image correction module comprises: A displacement correction unit performs the displacement corrections by using an affine transformation matrix, so that the starting point positions of the unselected images are corrected to be the same as the starting point positions of the target image. 8. An image processing method, which is used to process a plurality of images of a second resolution generated by enlarging a plurality of images of a first resolution, wherein one of the images of the second resolution is a target image, and the remaining The second resolution images for are multiple unselected images, including: Estimating multiple block displacements and multiple global displacements of the unselected images relative to the target image, so as to perform multiple displacement corrections on the unselected images, thereby generating multiple corrected images; judging whether the difference between the displacement of each block and the corresponding global displacement is greater than a threshold value, so as to generate a plurality of object movement pointers, wherein each of the object movement pointers is in a driving state and a blocking state 1. The driving state is that there is object movement in the block corresponding to the object movement pointer, and the blocking state is that there is no object movement in the block corresponding to the object movement pointer; multiple weights are set according to the moving pointers of the objects, multiple direction gradient operations are performed on each pixel of the target image and each pixel of the corrected images to generate multiple gradient differences, and these performing an arithmetic operation on each pixel of the target image and each pixel of the corrected images according to the object moving pointers and the gradient differences to generate a third resolution image, wherein the first The resolution of the three-resolution image is higher than the resolution of the second-resolution images, wherein when the object movement pointer is the driving state, then these weightings are set to zero, and when the object movement pointer is In the blocked state, the weights are set as the gradient differences. 9. The image processing method according to claim 8, wherein the arithmetic operation uses the weights to perform a weighted sum. 10. The image processing method according to claim 9, wherein when the difference between the block displacement and the corresponding global displacement is greater than the threshold value, the objects are driven to move the pointer, and when each of the When the difference between the block displacement and the corresponding global displacement is less than the threshold, the objects are blocked from moving the pointer. 11. The image processing method according to claim 9, wherein the direction gradient operations are performed for each pixel of the target image and each pixel of the corrected images to generate a plurality of direction gradient values, the These direction gradient values include a horizontal direction gradient value, a vertical direction gradient value and two diagonal direction gradient values. 12. The image processing method according to claim 11, wherein the step of performing the directional gradient calculations for each pixel of the target image and each pixel of the corrected images to generate the gradient differences comprises : Select the one with the largest numerical value among the direction gradient values as a maximum gradient value; selecting the smallest value among the direction gradient values as a minimum gradient value; and Make each gradient difference equal to the difference between the maximum gradient value and the minimum gradient value. 13. The image processing method according to claim 8, wherein the step of estimating the block displacements and the global displacements of the unselected images relative to the target image comprises: dividing the target image and the unselected images into multiple blocks; Estimating the block displacements of the unselected images relative to the target image; A plurality of global displacement estimations are performed according to the block displacements to generate the global displacements. 14. The image processing method according to claim 8, wherein the displacement corrections use an affine transformation matrix to correct the starting point positions of the unselected images to be the same as the starting point positions of the target image.