CN115205114A - High-resolution image splicing improved algorithm based on ORB (object-oriented bounding box) features - Google Patents
High-resolution image splicing improved algorithm based on ORB (object-oriented bounding box) features Download PDFInfo
- Publication number
- CN115205114A CN115205114A CN202210728085.XA CN202210728085A CN115205114A CN 115205114 A CN115205114 A CN 115205114A CN 202210728085 A CN202210728085 A CN 202210728085A CN 115205114 A CN115205114 A CN 115205114A
- Authority
- CN
- China
- Prior art keywords
- image
- matching
- algorithm
- points
- splicing
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000009466 transformation Effects 0.000 claims abstract description 24
- 239000011159 matrix material Substances 0.000 claims abstract description 20
- 230000004927 fusion Effects 0.000 claims abstract description 6
- 238000007499 fusion processing Methods 0.000 claims abstract description 4
- 230000006872 improvement Effects 0.000 claims description 7
- 238000005457 optimization Methods 0.000 claims description 4
- 230000002596 correlated effect Effects 0.000 claims description 3
- 230000001186 cumulative effect Effects 0.000 claims description 3
- 238000007500 overflow downdraw method Methods 0.000 claims description 3
- 238000013519 translation Methods 0.000 claims description 3
- 238000012544 monitoring process Methods 0.000 description 9
- 238000012545 processing Methods 0.000 description 8
- 238000005516 engineering process Methods 0.000 description 6
- 238000000034 method Methods 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 5
- 230000000007 visual effect Effects 0.000 description 5
- 238000011161 development Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 230000007547 defect Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 201000004569 Blindness Diseases 0.000 description 1
- 230000003190 augmentative effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000000875 corresponding effect Effects 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000012806 monitoring device Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
- G06T3/4038—Image mosaicing, e.g. composing plane images from plane sub-images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
- G06T3/4053—Scaling of whole images or parts thereof, e.g. expanding or contracting based on super-resolution, i.e. the output image resolution being higher than the sensor resolution
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/50—Image enhancement or restoration using two or more images, e.g. averaging or subtraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/757—Matching configurations of points or features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2200/00—Indexing scheme for image data processing or generation, in general
- G06T2200/32—Indexing scheme for image data processing or generation, in general involving image mosaicing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20212—Image combination
- G06T2207/20221—Image fusion; Image merging
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention provides an ORB feature-based high-resolution image stitching improved algorithm, which comprises the following steps: s1: extracting characteristic points of the spliced image; s2: fast rough matching is achieved on the feature points extracted in the S1; s3: optimizing the rough matching points in the S2, and eliminating wrong matching dot matrixes; s4: calculating relevant parameters of the homography transformation matrix, solving an image transformation matrix, and carrying out image splicing to obtain a primary spliced image; s5: and performing fusion processing on the overlapped area in the primarily spliced image by adopting a gradual-in and gradual-out weighting fusion algorithm, and removing the splicing trace of the image to obtain the spliced image. The algorithm provided by the invention improves the characteristic point matching step of the ORB algorithm, greatly improves the running speed, shortens the matching time, improves the matching accuracy to a certain extent, has good image splicing quality, and meets the use requirement of fast and accurate splicing of high-resolution image splicing.
Description
Technical Field
The invention belongs to the technical field of image processing, and relates to an improved high-resolution image stitching algorithm based on ORB characteristics.
Background
With the rapid development of socioeconomic and scientific technologies, various types of video monitoring equipment are installed and used in more and more places, such as some dangerous local areas and crowded public transportation places in factory enterprises. The traditional manual monitoring mode is gradually eliminated, and the video monitoring mode is selected to replace the laggard traditional monitoring mode. However, most of the existing video monitoring devices adopt fixed position shooting and collection, the visual field of a video picture obtained by a single camera device is limited, and potential safety hazards exist in a visual blind area. If shooting is carried out through the fisheye lens, the monitoring picture has a large distortion phenomenon, and the video monitoring effect is influenced. With the application of the image splicing technology, a plurality of camera devices can be combined to form a large-view-field and dead-angle-free video monitoring system.
The image stitching technology is always a very important and popular research direction in the field of computer vision and digital image processing technology, and in order to acquire a digital image with a larger visual field and higher resolution, two or more digital images with the same area are acquired by different visual angles or different sensors, a series of geometric transformation relations of graphs are carried out, and a single image with a larger visual field and higher quality is obtained by stitching. At present, the image stitching processing technology has gradually gained wide attention, popularization and application in a plurality of application development scenes such as a mobile virtual augmented reality image technology, remote sensing remote monitoring images, medical and health image data analysis, video image monitoring and the like.
Lowe D G et al summarize and publish a Scale Invariant Feature Transform (SIFT) algorithm, have milestone significance, can maintain good performance when the size and angle of an image change, have excellent splicing effect, and are rapidly popularized and applied. But the SIFT algorithm has long operation time and is not fast enough in matching speed. The method aims at the defects that the feature point description vector dimension of the SIFT algorithm is high, the algorithm complexity is high, the operation consumption is long, and the like. Bay et al propose an improved version of SIFT algorithm, SURF (speedup Robust Features) algorithm, which greatly increases the calculation and processing speed of SIFT algorithm by reducing and reducing the number of dimensions of the description of the feature points extracted from the image, but the matching accuracy is not high enough.
At present, the research on an improved image splicing algorithm usually focuses on the accuracy of image splicing and neglects the instantaneity and stability of image splicing processing. And the resolution of the image processed by the existing image splicing algorithm is still low, the rapid processing of the high-resolution image cannot be completed, and the resolution of the acquired image becomes higher and higher along with the rapid development of the image acquisition equipment. However, the existing image mosaic algorithm is slow in processing speed, can only process images with low resolution, cannot meet the requirement of high resolution in practical application, and still has great limitation in application and popularization.
Disclosure of Invention
The invention aims to provide an improved high-resolution image stitching algorithm based on ORB characteristics, which comprises the following steps:
s1: extracting characteristic points of the spliced image;
s2: fast rough matching is achieved on the feature points extracted in the S1;
s3: optimizing the rough matching points in the S2, and eliminating wrong matching dot matrixes;
s4: calculating relevant parameters of the homography transformation matrix, solving an image transformation matrix, and carrying out image splicing to obtain a primary spliced image;
s5: and performing fusion processing on the overlapped area in the primarily spliced image by adopting a gradual-in and gradual-out weighting fusion algorithm, and removing the splicing trace of the image to obtain the spliced image.
The improved high-resolution image stitching algorithm based on the ORB features also has the characteristics that the FAST algorithm is adopted in the S1 to extract the feature points of the stitched image, and then the BRIEF algorithm description is carried out on the extracted feature points.
The invention provides an improved algorithm for splicing high-resolution images based on ORB characteristics, which is also characterized in that S2 comprises the following steps:
s2.1: and (3) judging description similarity by using the Hamming distance, and calculating the cumulative sum of the bitwise AND operation of the two description strings as the Hamming distance D (X, Y):
wherein the value of the Hamming distance is positively correlated with the correlation of the description string,
s2.2: and carrying out coarse matching on the characteristic points acquired in the step S1 by using a pattern matcher.
The ORB feature-based high-resolution image stitching improved algorithm further has the characteristic that a PROSAC algorithm is adopted in S3 to purify and optimize the feature points.
The invention provides an ORB feature-based high-resolution image stitching improved algorithm, which is further characterized in that S3 comprises the following steps:
s3.1: initializing relevant parameters and inputting maximum iteration number N max Determining error limit E of interior point lim Setting an interior point number threshold T num ；
S3.2: according to the result of the ratio of the nearest neighbor to the next nearest neighbor of the coarse matching point pairs as the matching quality, performing descending order arrangement, selecting m matching point pairs before the arrangement of the matching quality, selecting at least 4 groups of point pairs from the m matching point pairs, calculating a homography transformation matrix, calculating the errors of the rest matching point pairs and the projection points, and comparing the errors with the errors E lim If the error is less than E lim The point is the interior point, otherwise, it is the exterior point. Traversing all the matching point pairs, and calculating the number of the inner points;
s3.3: judging whether the iteration number is less than or equal to the maximum iteration number N max If the condition is met, the execution is continued; if the iteration number is larger than the maximum iteration number N max Outputting a group of matching point pair sets with the most interior points in statistics;
s3.4: comparing the number of inliers to a threshold T num If the number of interior points is less than the threshold value T num If so, adding 1 to the iteration times, and repeatedly executing the steps S3.2-S3.3; otherwise, outputting the current inner point set to complete the optimization of the matching point pair.
The ORB feature-based high-resolution image stitching improved algorithm further has the characteristic that S4 obtains the optimal matching feature point set f of two images to be stitched after the mismatching point pairs are removed a (x,y),f b (x, y), calculating relevant parameters of the image homography transformation matrix, performing image splicing after obtaining the image transformation matrix, and obtaining a primary spliced image, wherein the spatial homography transformation relation of the two image characteristic points is as follows:
wherein: h is a total of 0 、h 1 、h 3 And h 4 Jointly representing the graphic rotation angle and the scale, h 2 And h 5 Respectively representing the amount of translation of the graph in the x-direction and the y-direction, h 6 And h 7 Respectively representing the amount of deformation of the figure in the x-direction and the y-direction.
The improved algorithm for splicing the high-resolution images based on the ORB features further has the following characteristics that in the step S5, splicing traces are removed through a fade-in fade-out weighted fusion method:
for two images f 1 (x,y),f 2 (x, y) is calculated as follows:
in the formula, the weighted weight value
Both are related to the width of the overlapping area of the two images, x 1 And x 2 Is the abscissa of the pixel point at the left and right edges of the overlap region, where x 1 <x 2 ，x i Is the abscissa of the pixel point to be fused in the overlapping region.
The invention has the beneficial effects that:
the improved ORB feature-based high-resolution image stitching algorithm improves the feature point matching step of the ORB algorithm through the PAOSAC algorithm, greatly improves the running speed, shortens the matching time, improves the matching accuracy to a certain extent, has good image stitching quality, and meets the use requirement of quick and accurate stitching of high-resolution image stitching.
Drawings
Fig. 1 is a flowchart of an ORB-feature-based high-resolution image stitching improvement algorithm provided by the present invention.
Detailed Description
The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings and specific embodiments. It is to be understood that the embodiments described are merely exemplary of the invention and that, based on the embodiments described herein, all other embodiments obtained by persons skilled in the art without making any creative effort fall within the scope of the present invention.
As shown in fig. 1, the present invention provides an improved algorithm for stitching high-resolution images based on ORB features, which includes the following steps:
s1: extracting characteristic points of the spliced image;
s2: fast rough matching is achieved on the feature points extracted in the S1;
s3: optimizing the rough matching points in the S2, and eliminating wrong matching dot matrixes;
s4: calculating relevant parameters of the homography transformation matrix, solving an image transformation matrix, and carrying out image splicing to obtain a primary spliced image;
s5: and performing fusion processing on the overlapped area in the primarily spliced image by adopting a gradual-in and gradual-out weighting fusion algorithm, and removing the splicing trace of the image to obtain the spliced image.
In the embodiment, the PAOSAC algorithm is used for improving the characteristic point matching step of the ORB algorithm, the running speed is greatly improved, the matching time is shortened, the matching accuracy is improved to a certain extent, the image splicing quality is good, and the use requirement of quick and accurate splicing of high-resolution image splicing is met.
In some embodiments, in S1, a FAST algorithm is used to extract feature points of the stitched image, and then a BRIEF algorithm description is performed on the extracted feature points.
In the embodiment, the FAST algorithm is adopted to detect the feature points of the image, and then the BRIEF algorithm description is carried out on the detected feature points, so that the advantages of the FAST algorithm and the BRIEF algorithm description are combined, the defects of the BRIEF algorithm description are improved, the interference of image noise to the BRIEF algorithm is solved, the BRIEF algorithm has good rotation invariance, and the rapid and accurate feature point extraction of the digital image can be realized. The FAST algorithm can quickly detect the characteristic points, in a digital image, a circular neighborhood region is selected by comparing the gray value of a certain point pixel with the gray value of the certain neighborhood region, generally taking the pixel as the center, and when the difference between the gray value of enough pixels in the region and the gray value of the pixel is greater than a threshold value, the pixel is judged as the characteristic point. Meanwhile, since the speed of the FAST algorithm is very surprising, the ORB adopts the improved FAST to solve the scale invariance, and the practicality of the ORB is mainly in the application requiring higher speed, so that a foundation is provided for the FAST processing of high-resolution images.
In some embodiments, the S2 includes the following steps:
s2.1: and (3) judging description similarity by using the Hamming distance, and calculating the cumulative sum of the bitwise AND operation of the two description strings as the Hamming distance D (X, Y):
wherein the value of the Hamming distance is positively correlated with the descriptive string correlation,
s2.2: and carrying out coarse matching on the characteristic points acquired in the step S1 by using a pattern matcher.
For rough matching of feature points, the traditional algorithm adopts nearest neighbor matching feature point pairs to express correct matching, a large number of feature matching pairs are introduced, many mismatching exists, and the matching accuracy is low. In the above embodiment, hamming distance (Hamming distance) is used to determine similarity (XOR operation) between descriptors, matching of key points will become a binary classification problem, an accumulated sum of bitwise and operated values of two description strings is calculated, a smaller Hamming distance value indicates a higher correlation between the two description strings, and otherwise, a larger difference between the two description strings is obtained. And finally, carrying out coarse matching on all the feature points by using a pattern matcher.
In some embodiments, the S3 adopts a PROSAC algorithm to refine and optimize the feature points.
The PROSAC algorithm is an algorithm for compensating randomness of the RANSAC algorithm, and has higher robustness and computational efficiency than the RANSAC algorithm. In the embodiment, a PROSAC algorithm is adopted to optimize the image matching feature point pairs, the point pairs in the sample set are sorted according to the matching quality, the sample at the front position is selected as the hypothesis set, and a subset with higher matching probability is generated, i.e. sampling calculation is performed in the coarse matching subset with high quality, so that the blindness of selecting the initial hypothesis set of the sample is avoided, and the algorithm operation efficiency is improved to a great extent. Compared with the RANSAC algorithm, the method has the advantages that the iteration times are greatly reduced, the calculation speed is increased, and meanwhile, the matching accuracy is improved.
In some embodiments, the S3 includes the following steps:
s3.1: initializing relevant parameters and inputting maximum iteration number N max Determining error limit E of interior point lim Setting a threshold value T of the number of interior points num ；
S3.2: according to the result of the ratio of the nearest neighbor to the next nearest neighbor of the rough matching point pairs as the matching quality, performing descending order arrangement, selecting m matching point pairs before the arrangement of the matching quality, selecting at least 4 groups of point pairs randomly, calculating a homography transformation matrix, calculating the errors of the rest matching points and the projection points, and comparing the errors with the errors E lim If the error is less than E lim This point is the inner point, otherwise, it is the outer point. Traversing all the matching point pairs, and calculating the number of the inner points;
s3.3: judging whether the iteration number is less than or equal to the maximum iteration number N max If the condition is met, the execution is continued; if the number of iterationsGreater than the maximum number of iterations N max Outputting a group of matching point pair sets with the most interior points in statistics;
s3.4: comparing the number of inliers to a threshold T num If the number of interior points is less than the threshold value T num If so, adding 1 to the iteration number, and repeatedly executing the steps S3.2-S3.3; otherwise, outputting the current inner point set to complete the optimization of the matching point pair.
In some embodiments, after the mismatching point pairs are removed, the S4 obtains an optimal matching feature point set f of the two images to be stitched a (x,y),f b (x, y), calculating relevant parameters of the image homography transformation matrix, performing image splicing after obtaining the image transformation matrix, and obtaining a primary spliced image, wherein the spatial homography transformation relation of the two image characteristic points is as follows:
wherein: h is a total of 0 、h 1 、h 3 And
h 4 jointly representing the graphic rotation angle and the scale, h 2 And h 5 Respectively representing the amount of translation of the graph in the x-direction and the y-direction, h 6 And h 7 Respectively representing the amount of deformation of the figure in the x-direction and the y-direction.
In order to determine the homography transformation matrix H, 8 undetermined parameters in the equation need to be solved. At least 4 groups of matching feature point pairs are selected from the optimal matching feature point set of the two images, calculation is carried out through the formula, 8 parameters are solved, and the optimal solution can be calculated through selecting multiple groups to obtain a transformation matrix H. And multiplying the transformation matrix H by the image to be spliced to linearly add the result image and the reference image to obtain a spliced image.
In some embodiments, the acquired digital images may be acquired at different times or by different image sensors, and the shooting angle and the integration time setting of the digital images may also be different, so that the images may have differences in pixel gray levels, and after the images are spliced, a blurred ghost phenomenon may occur in an image overlapping area, which requires a certain image fusion algorithm to eliminate splicing marks. And in the S5, removing splicing traces by a gradual-in gradual-out weighting fusion method:
for two images f 1 (x,y),f 2 (x, y) is calculated as follows:
in the formula, the weighted weight value
Both are related to the width of the overlapping area of the two images, x 1 And x 2 Is the abscissa of the pixel point at the left and right edges of the overlap region, where x 1 <x 2 ，x i Is the abscissa of the pixel point to be fused in the overlapping region.
In the embodiment, after the gradual-in and gradual-out weighted fusion, the gray value of the overlapped area is calculated according to the corresponding weight, so that the gray difference is reduced, the splicing trace can be effectively eliminated, the spliced image is smoother and more natural, and the imaging quality is better.
The image mosaic is performed by using the SIFT algorithm, the SURF algorithm and the method provided by the embodiment, and the mosaic result is as the following table 1 and table 2, wherein the resolutions of the two groups of pictures are respectively (a) 1280 pixels × 720 pixels (720 p) and (b) 1920 pixels × 1080 pixels (1080 p), and the image format is JPG.
TABLE 1 comparison of image stitching speeds for different algorithms
As can be seen from Table 1, the splicing method provided by the invention has the shortest feature point extraction time, and the algorithm calculation speed of the embodiment is much higher than that of the SIFT algorithm and the SURF algorithm. Because the PROSAC algorithm is adopted to optimize the matching of the image feature points, the iteration times are reduced compared with RANSAC, the operation efficiency is higher, the feature point matching time is the shortest, and compared with the SIFT algorithm and the SURF algorithm, the speed of the algorithm of the embodiment is respectively improved by about 24 times and 8 times.
TABLE 2 comparison of accuracy of image stitching for different algorithms
As can be seen from table 2, the matching accuracy CMR of the embodiment is higher than the SIFT algorithm and the SURF algorithm, and it can be seen that the PROSAC algorithm can remove more error matches and significantly improve the CMR compared to the RANSAC algorithm. And secondly, the embodiment adopts a PROSAC algorithm for optimization, so that the method has higher robustness and improves the matching accuracy. The root mean square error RMSE of the embodiment is lower than that of the SIFT algorithm and the SURF algorithm, and the registration accuracy of the embodiment is higher.
In conclusion, compared with other algorithms, the algorithm provided by the invention has the advantages that the operation speed is greatly improved, the matching accuracy is improved to a certain extent, the image splicing quality is good, and the requirement for quickly and accurately splicing the high-resolution images can be met.
The above description is intended to be illustrative of the preferred embodiment of the present invention and should not be taken as limiting the invention, but rather, the intention is to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the invention. The above description is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, several modifications and variations can be made without departing from the technical principle of the present invention, and these modifications and variations should also be regarded as the protection scope of the present invention.
Claims (7)
1. An improved algorithm for splicing high-resolution images based on ORB features is characterized by comprising the following steps:
s1: extracting characteristic points of the spliced image;
s2: fast rough matching is achieved on the feature points extracted in the S1;
s3: optimizing the rough matching points in the S2, and eliminating wrong matching dot matrixes;
s4: calculating relevant parameters of the homography transformation matrix, solving an image transformation matrix, and carrying out image splicing to obtain a primary spliced image;
s5: and performing fusion processing on the overlapped area in the primarily spliced image by adopting a gradual-in and gradual-out weighting fusion algorithm, and removing the splicing trace of the image to obtain the spliced image.
2. The ORB feature-based high-resolution image stitching improvement algorithm according to claim 1, wherein in S1, a FAST algorithm is adopted to extract feature points of the stitched image, and then the extracted feature points are subjected to BRIEF algorithm description.
3. The ORB feature based high resolution image stitching improvement algorithm of claim 1, wherein the S2 comprises the steps of:
s2.1: and (3) judging description similarity by using the Hamming distance, and calculating the cumulative sum of the bitwise AND operation of the two description strings as the Hamming distance D (X, Y):
wherein the value of the Hamming distance is positively correlated with the descriptive string correlation,
s2.2: and carrying out coarse matching on the characteristic points acquired in the step S1 by using a pattern matcher.
4. The ORB feature-based high-resolution image stitching improvement algorithm according to claim 1, wherein a PROSAC algorithm is adopted in S3 to refine and optimize feature points.
5. The ORB feature-based high resolution image stitching improvement algorithm according to claim 4, wherein the S3 comprises the following steps:
s3.1: initializing relevant parameters and inputting maximum iteration number N max Determining error limit E of interior points lim Setting a threshold value T of the number of interior points num ；
S3.2: according to the result of the ratio of the nearest neighbor to the next nearest neighbor of the rough matching point pairs as the matching quality, performing descending order arrangement, selecting m matching point pairs before the arrangement of the matching quality, selecting at least 4 groups of point pairs randomly, calculating a homography transformation matrix, calculating the errors of the rest matching points and the projection points, and comparing the errors with the errors E lim If the error is less than E lim The point is the interior point, otherwise, it is the exterior point. Traversing all the matching point pairs, and calculating the number of the inner points;
s3.3: judging whether the iteration number is less than or equal to the maximum iteration number N max If the condition is met, the execution is continued; if the iteration number is larger than the maximum iteration number N max Outputting a group of matching point pair sets with the most interior points in statistics;
s3.4: comparing the number of interior points to a threshold T num If the number of interior points is less than the threshold value T num If so, adding 1 to the iteration times, and repeatedly executing the steps S3.2-S3.3; otherwise, outputting the current inner point set to complete the optimization of the matching point pair.
6. The ORB feature-based high-resolution image stitching improvement algorithm according to claim 1, wherein S4 is used for obtaining an optimal matching feature point set f of two images to be stitched after removing the mismatching point pairs a (x,y),f b (x, y), calculating relevant parameters of the image homography transformation matrix, performing image splicing after obtaining the image transformation matrix, and obtaining a primary spliced image, wherein the spatial homography transformation relation of the two image characteristic points is as follows:
wherein: h is 0 、h 1 、h 3 And h 4 Jointly representing the graphic rotation angle and the scale, h 2 And h 5 Representing the translation of the figure in the x-direction and y-direction, respectivelyThe amounts, h6 and h7, represent the deformation amounts of the figure in the x-direction and the y-direction, respectively.
7. The ORB feature-based high-resolution image stitching improvement algorithm according to claim 1, wherein the stitching trace is removed in S5 by a fade-in and fade-out weighted fusion method:
for two images f 1 (x,y),f 2 (x, y) is calculated as follows:
in the formula, the weighted weight value
Both are related to the width of the overlapping area of the two images, x 1 And x 2 Is the abscissa of the pixel point at the left and right edges of the overlap region, where x 1 <x 2 ，x i Is the abscissa of the pixel point to be fused in the overlapping region.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210728085.XA CN115205114A (en) | 2022-06-24 | 2022-06-24 | High-resolution image splicing improved algorithm based on ORB (object-oriented bounding box) features |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210728085.XA CN115205114A (en) | 2022-06-24 | 2022-06-24 | High-resolution image splicing improved algorithm based on ORB (object-oriented bounding box) features |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115205114A true CN115205114A (en) | 2022-10-18 |
Family
ID=83577924
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210728085.XA Pending CN115205114A (en) | 2022-06-24 | 2022-06-24 | High-resolution image splicing improved algorithm based on ORB (object-oriented bounding box) features |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115205114A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115883988A (en) * | 2023-02-17 | 2023-03-31 | 南昌航天广信科技有限责任公司 | Video image splicing method and system, electronic equipment and storage medium |
| CN116309036A (en) * | 2022-10-27 | 2023-06-23 | 杭州图谱光电科技有限公司 | Microscopic image real-time stitching method based on template matching and optical flow method |
| CN116757936A (en) * | 2023-08-22 | 2023-09-15 | 佗道医疗科技有限公司 | Image matching relation acquisition method and image stitching method thereof |
-
2022
- 2022-06-24 CN CN202210728085.XA patent/CN115205114A/en active Pending
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116309036A (en) * | 2022-10-27 | 2023-06-23 | 杭州图谱光电科技有限公司 | Microscopic image real-time stitching method based on template matching and optical flow method |
| CN116309036B (en) * | 2022-10-27 | 2023-12-29 | 杭州图谱光电科技有限公司 | Microscopic image real-time stitching method based on template matching and optical flow method |
| CN115883988A (en) * | 2023-02-17 | 2023-03-31 | 南昌航天广信科技有限责任公司 | Video image splicing method and system, electronic equipment and storage medium |
| CN116757936A (en) * | 2023-08-22 | 2023-09-15 | 佗道医疗科技有限公司 | Image matching relation acquisition method and image stitching method thereof |
| CN116757936B (en) * | 2023-08-22 | 2023-11-07 | 佗道医疗科技有限公司 | Image matching relation acquisition method and image stitching method thereof |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107844779B (en) | Video key frame extraction method | |
| CN115205114A (en) | High-resolution image splicing improved algorithm based on ORB (object-oriented bounding box) features | |
| CN108108746B (en) | License plate character recognition method based on Caffe deep learning framework | |
| CN106683119B (en) | Moving vehicle detection method based on aerial video image | |
| CN108364272A (en) | A kind of high-performance Infrared-Visible fusion detection method | |
| CN110020995B (en) | Image splicing method for complex images | |
| CN107688829A (en) | A kind of identifying system and recognition methods based on SVMs | |
| CN103902953B (en) | A kind of screen detecting system and method | |
| CN111553845B (en) | Quick image stitching method based on optimized three-dimensional reconstruction | |
| CN103841298A (en) | Video image stabilization method based on color constant and geometry invariant features | |
| CN109376641A (en) | A kind of moving vehicle detection method based on unmanned plane video | |
| CN114529593A (en) | Infrared and visible light image registration method, system, equipment and image processing terminal | |
| CN112614167A (en) | Rock slice image alignment method combining single-polarization and orthogonal-polarization images | |
| CN108830283B (en) | Image feature point matching method | |
| CN112037109A (en) | Improved image watermarking method and system based on saliency target detection | |
| CN113052170A (en) | Small target license plate recognition method under unconstrained scene | |
| CN110516731B (en) | Visual odometer feature point detection method and system based on deep learning | |
| Lecca et al. | Comprehensive evaluation of image enhancement for unsupervised image description and matching | |
| Zhang et al. | Multi-scale segmentation strategies in PRNU-based image tampering localization | |
| CN111127353A (en) | High-dynamic image ghost removing method based on block registration and matching | |
| CN115456870A (en) | Multi-image splicing method based on external parameter estimation | |
| CN106845555A (en) | Image matching method and image matching apparatus based on Bayer format | |
| CN113436251A (en) | Pose estimation system and method based on improved YOLO6D algorithm | |
| CN116403115B (en) | Large-format remote sensing image target detection method | |
| CN112184533A (en) | Watermark synchronization method based on SIFT feature point matching |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |