CN107093166B - The seamless joint method of low coincidence factor micro-image - Google Patents
The seamless joint method of low coincidence factor micro-image Download PDFInfo
- Publication number
- CN107093166B CN107093166B CN201710212589.5A CN201710212589A CN107093166B CN 107093166 B CN107093166 B CN 107093166B CN 201710212589 A CN201710212589 A CN 201710212589A CN 107093166 B CN107093166 B CN 107093166B
- Authority
- CN
- China
- Prior art keywords
- image
- matching
- pair
- spliced
- characteristic point
- 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformation in the plane of the image
- G06T3/40—Scaling the whole image or part thereof
- G06T3/4038—Scaling the whole image or part thereof for image mosaicing, i.e. plane images composed of plane sub-images
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10056—Microscopic image
-
- 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
Abstract
The present invention discloses a kind of seamless joint method of low coincidence factor micro-image, the described method comprises the following steps: choosing the region of interest ROI with reference to figure and figure to be spliced, pre-processes to ROI；The SURF characteristic point for extracting ROI, slightly matches characteristic point using FLANN method；The excessive matching pair of space length is rejected according to the deviation range of two images；The discrete distribution of the space length of the remaining matching pair of statistics further rejects error hiding pair to be maximally distributed section corresponding to probability as critical field, if matching is to deficiency, abovementioned steps are returned, feature extraction parameter and match parameter are modified, to obtain more matchings pair；According to remainder matching pair, the positional relationship of two images is calculated；It treats spliced map and carries out brightness adjustment, weighted mean method is taken to merge image.This method is based on improved SURF Characteristic points match method, characteristic matching precision and efficiency is effectively improved in the case where image low Duplication, to improve the quality and speed of image mosaic.
Description
Technical field
It is the present invention relates to digital image processing techniques field, in particular to seamless under the low coincidence factor of a kind of adjacent micro-image
The method of splicing.
Background technique
The problem of microscope inspection observation of biological sample always exists very time-consuming and inefficiency, current full-automatic microscopy
Platform gradually replaces original cervical arthroplasty, this cost saved manpower.In order to obtain the micro-image of big visual field large scene,
It needs to splice the image under more scenes.If full-automatic microscopy machine needs the big field-of-view image under N number of different low coverages, this
It needs effectively this N number of Microscopic Image Mosaicing at a secondary seamless image, in case subsequent automatic classification, the knowledge for realizing biological sample
Not with counting.
Image mosaic is applied to commercial measurement, and core technology is image registration techniques, including template method for registering, phase
Degree of correlation method and characteristic matching method.These methods are registration basis with the overlapping region of adjacent image, dependent on image itself
Feature is registrated, therefore splicing precision is not high, and calculation amount is very big, relatively high to the overlap proportion requirement of overlapping region, is had
Even to reach 50% or more overlap proportion.When measurement object lens multiple is higher, measurement visual field itself is smaller
In the case of, big overlap proportion seriously constrains expansion of the splicing to measurement visual field area, especially when overlapping region is not bright
When aobvious feature, or even it will appear the case where can not matching.
Summary of the invention
The purpose of the present invention is to provide a kind of seamless joint methods of low coincidence factor micro-image, and this method can be effective
Improve the precision and speed of image mosaic under more scenes in ground.
Realize the technical scheme is that
A kind of seamless joint method of low coincidence factor micro-image, method includes the following steps:
Step 1: choosing the area-of-interest (ROI) with reference to figure and figure to be spliced, ROI is pre-processed；
Step 2: SURF (Speed-up robust feature) characteristic point of ROI is extracted, using FLANN (Fast
Library Approximate Nearest Neighbors) method slightly matches characteristic point；
Step 3: the excessive matching pair of space length is rejected according to the deviation range of two images；
Step 4: the discrete distribution of the space length of the remaining matching pair of statistics is to be maximally distributed section corresponding to probability
Critical field further rejects error hiding pair；If matching returns to abovementioned steps to deficiency, feature extraction parameter and matching are modified
Parameter, to obtain more matchings pair；
Step 5: according to remainder matching pair, calculating the positional relationship of two images；
Step 6: treating spliced map and carry out brightness adjustment, weighted mean method is taken to merge image；Wherein:
Coincidence factor of the low coincidence factor between adjacent image is between 3%~11%.
The step 1 specifically:
Read reference pictureI 0(x, y) and image to be splicedI 1(x, y), choose facing comprising lap in two width figures
Near field is as region of interest ROI, using the noise spot in median filtering method removal ROI, wherein filter window size is set
It is set to 3 × 3；
Using the method for histogram equalization, expand its dynamic range, to enhance the contrast of image.
The step 2 specifically:
The SURF characteristic point in ROI is extracted respectively, and each characteristic point is described with an isometric feature vector, extracted
Feature point set is denoted as F0 and F1 respectively；
Using FLANN method, for each characteristic point in F0, found out in F1 therewith feature vector apart from the smallest K
It is a, K=2, if the ratio of minimum range and time small distance is less than certain threshold value, then it is assumed that point corresponding to minimum range is
Match；Otherwise it is assumed that this feature point in F0 is not matched characteristic point in F1；Wherein, certain threshold value initial value
It is set as 0.5.
The step 3 specifically:
According to the feature of multiple acquired image pattern, image Duplication is between 3% ~ 11%, the offset of vertical direction
Within 5%, larger boundary 11% and 5% is taken, obtains corresponding boundary shifts amount, as thresholding, to every in thick matching result
To match point, judge whether its space length meets the threshold condition, satisfaction then retains；It is unsatisfactory for, is considered as error hiding and is picked
It removes.
The step 4 specifically:
The space length of matching double points is divided into the equal section of multiple step-lengths, by space length discretization；To upper one
All matchings pair remained are walked, calculate matching to the discrete distribution situation of space length；If secondary maximum probability and maximum
The ratio of probability is less than certain threshold value T, then choosing section corresponding to maximum probability is critical field；If the ratio is greater than certain threshold
Value, and secondary maximum probability is adjacent with section corresponding to maximum probability, then chooses the interlude of the two adjacent intervals as standard
Range；Other situations, then return step 2, change the extracting parameter and the matched parameter of FLANN of SURF characteristic point, to obtain
More matching double points；Successively judge whether the space length for the matching double points that every a pair remains meets the critical field,
The conduct for meeting condition is correctly matched to reservation, and the ungratified error hiding that is considered as is to being removed；Judge finally remained
Whether reach certain amount with logarithm, if lazy weight, return step 2 changes the extracting parameter of SURF characteristic point, and
The matched parameter of FLANN, to obtain more matching double points；Wherein, certain threshold value T value range is 0.4 ~ 0.6.
The step 5 specifically:
To the matching double points finally remained, according to the spatial position of match point, reference picture is sought using the method for average
With the spatial offset of image to be spliced, positional relationship between the two is determined.
The step 6 specifically:
According to the relative positional relationship for the two images that previous step obtains, the real overlapping region of two images is determined；
The RGB color mean value for calculating separately overlapping region between reference picture and image to be spliced, obtains each of reference picture
Each color component the mean value R1, G1, B1 of color component mean value R0, G0, B0 and image to be spliced；
The color mean value of image to be spliced subtracts the color mean value of reference picture, color difference Dr, Dg, Db is obtained, with this
Overall brightness difference as two images；
Each pixel of stitching image is treated, each color component adds Dr, Dg, Db respectively, by its luminance level tune
It is whole to consistent with reference picture；
Each pixel of image overlapping region to be spliced to reference picture and by brightness adjustment, it is in place according to its institute
It sets, different weights is set, with the progressive weighted average method gradually gone out, overlapping region is merged, non-overlap area is then direct
Replicate original image.
The beneficial effects of the present invention are: the seamless spliced side of micro-image under a kind of low coincidence factor proposed by the invention
Method uses improved SURF Characteristic points match method in the case where adjacent image low Duplication, effectively extracts characteristics of image simultaneously
The efficiency and precision of characteristic matching are improved, to improve the quality and splicing speed of image mosaic.
Detailed description of the invention
Fig. 1 is the image sequence schematic diagram of acquisition provided by the invention；
Fig. 2 is flow chart of the invention；
Fig. 3 is the improved image split-joint method flow chart based on characteristic point of the present invention；
Fig. 4 is influence schematic diagram of the image enhancement provided by the invention to feature point extraction effect；Fig. 4 (a) is not carry out
The characteristic point that image enhancement uses SURF algorithm to extract；Fig. 4 (b) is the characteristic point for using SURF algorithm to extract image enhancement；
Fig. 5 is method provided by the invention and the effect contrast figure of other matching process；Fig. 5 (a) is that FLANN is slightly matched
Effect picture, Fig. 5 (b) be using RANSAC algorithm to Fig. 5 (a) carry out error hiding rejecting after effect picture, Fig. 5 (c) be use
Improved characteristic point matching method rejects the effect picture after the error hiding of part to Fig. 5 (a)；
Fig. 6 is the effect picture provided by the invention promoted using brightness adjustment to fusion mass；Fig. 6 (a) be reference picture,
6 (b) be image to be spliced, 6 (c) be do not carry out brightness adjustment directly to after image co-registration effect picture, 6 (d) be to carry out brightness
Adjustment is then to the effect picture after image co-registration；
Fig. 7 is the acquisition image that 4 × 4 are directed to using image split-joint method of the invention, the effect picture spliced；Fig. 7
(a) it is 4 × 4 original dermal harvest image, 7 (b) is original 4 × 4 tubercle bacillus acquisition image；Fig. 7 (c) is splicing
Rear skin image, 7 (d) are spliced tubercle bacillus images.
Specific embodiment
Illustrate technological means, technological improvement and beneficial effect of the present invention in order to be more clearly understood, ties below
Closing attached drawing, the present invention will be described in detail.
Automatic microscopy platform mostly uses the movement of step motor control objective table, since stepper motor is in the spy of precision aspect
Property and existing return difference problem during the motion, acquired image is not ensured that in X-axis and Y-axis after continuous translation
Absolutely alignment, always slightly deviation, Fig. 1 are the image sequence schematic diagrames that automatic microscopy machine obtains；But stepper motor generally controls
Precision or very high, therefore for the Microscopic Image Mosaicing problem obtained under automatic microscopy machine, one kind provided by the present invention
Based on the micro-image seamless joint method under low coincidence factor, referring to fig. 2 and Fig. 3, including the following steps:
S101: the area-of-interest (ROI) with reference to figure and figure to be spliced is chosen, ROI is pre-processed.
The step specifically:
Read reference pictureI 0(x, y) and image to be splicedI 1(x, y)；
It choosesI 0(x, y) andI 1(x, y) in adjacent domain including overlay region as area-of-interest (ROI), note
ForI R0 WithI R1 ；
Using median filtering method pairI R0 WithI R1 It is filtered to remove abnormal point, wherein the window size of median filtering is set
3 × 3 are set to, treated, and image is denoted asI R0 ’ WithI R1 ’ ；
It is rightI R0 ’ WithI R1 ’ , using the method for histogram equalization, expand its dynamic range, to enhance the contrast of image,
Treated, and image is denoted asI R0 ’’ WithI R1 ’’ 。
S102: extracting the SURF characteristic point of ROI, is slightly matched using FLANN method to characteristic point.
The step specifically:
It extracts respectivelyI R0 ’’ WithI R1 ’’ In SURF characteristic point, the feature point set extracted is denoted as F respectively0And F1；
Using FLANN method, for each characteristic point in F0, found out in F1 therewith feature vector apart from the smallest K
(K can be set as 2) a point, if the ratio of minimum range and time small distance is less than certain threshold value (initial value is set as 0.5), then it is assumed that most
Point corresponding to small distance is matched；Otherwise it is assumed that this feature point in F0 is not matched characteristic point in F1；
Matching is denoted as M to set1。
S103: the excessive matching pair of space length is rejected according to the deviation range of two images.
The step specifically:
According to the feature of multiple acquired image pattern, image Duplication is between 3% ~ 11%, the offset of vertical direction
Within 5%, larger boundary 11% and 5% is taken, corresponding boundary shifts amount is obtained, as thresholding, to M1In each pair of match point,
Judge whether its space length meets the threshold condition, satisfaction then retains；It is unsatisfactory for, is considered as error hiding and is removed.It remains
The set being composed of be denoted as M2。
S104: the discrete distribution of the space length of remaining matching pair is counted, is to be maximally distributed section corresponding to probability
Critical field further rejects error hiding pair；If matching returns to S102 to deficiency, feature extraction parameter and match parameter are modified,
To obtain more matchings pair.
The step specifically:
By M2The space length of middle matching double points is divided into the equal section d of multiple step-lengths1, d2, d3,…dn, by space
Apart from discretization；
To M2In all matching double points, calculate the discrete distribution situation of its space length, that is, fall in the general of each section
Rate；
Maximum probability is denoted as P1, secondary maximum probability is denoted as P2, P2With P1Ratio be denoted as Ratio；
If the value of Ratio is less than certain threshold value (value range is 0.4 ~ 0.6), P is chosen1Corresponding section is as standard
Section ds；If Ratio is greater than certain threshold value, and P1With P2Corresponding section is adjacent, then the interlude for choosing the two sections is made
For standard section ds；Other situations, then return step S102, extracting parameter and the FLANN for changing SURF characteristic point are matched
Parameter, to obtain more matching double points；
Successively judge M2In the space lengths of every a pair of of matching double points whether belong to standard section ds, belong to dsConduct just
Really to reservation, what is be not belonging to is considered as error hiding to being removed for matching, and the set of the matching pair remained is denoted as M3；
Judge M3In matching logarithm whether reach certain amount, if lazy weight, it is special to change SURF by return step S102
The extracting parameter and the matched parameter of FLANN of point are levied, to obtain more matching double points.
S105: according to the matching double points finally remained, the positional relationship between reference picture and image to be spliced is calculated.
The step specifically:
To M3In matching double points reference picture and to be spliced is sought using the method for average according to the spatial position of match point
The spatial offset of image determines the positional relationship between reference picture and matching image.
S106: it treats stitching image and carries out brightness adjustment, weighted mean method is taken to carry out splicing fusion to image.
The step specifically:
According to the relative positional relationship for the two images that previous step obtains, the real overlapping region of two images is determined,I 0
(x, y) in overlay region be denoted as r0, I 1(x, y) in overlay region be denoted as r1；
Calculate separately r0And r1RGB color mean value, obtain r0Each color component mean value R0, G0, B0And r1Each face
Colouring component mean value R1, G1, B1；
R1, G1, B1With R0, G0, B0It is poor to make respectively, color difference Dr, Dg, Db is obtained, in this, as the whole of two images
Body luminance difference；
Treat stitching imageI 1(x, y) each pixel, each color component adds Dr, Dg, Db respectively, as a result remembers
ForI 1 ’(x, y), it willI 1(x, y) luminance level be adjusted to and reference pictureI 0(x, y) consistent；
It is rightI 0(x, y) andI 1 ’(x, y) different power is arranged according to its position in each pixel of overlapping region
Weight merges overlapping region with the progressive weighted average method gradually gone out, and original image is then directly replicated in non-overlap area, is spelled
Map interlinking pictureI mosaic。
When table 1 is spent by application the method for the present invention provided by the invention and application RANSAC method raising matching precision
Between comparison diagram.
Table 1
| Smart matching process | RANSAC (s) | Method (s) provided by the invention |
| Skin Cell figure | 0.140 | 0.027 |
| Tubercle bacillus figure | 0.513 | 0.021 |
In conclusion under a kind of micro-image proposed by the invention low coincidence factor image seamless joining method, matching
It is based on SURF characteristic point in quasi- method, the contrast of image is further improved by image enchancing method, makes the feature of image more
For protrusion, the characteristic point extracted is slightly matched using FLANN algorithm, then using based on space length between match point
The method of feature gradually rejects the Mismatching point pair in thick matching result；If the correct matching double points remained are insufficient, can lead to
The extracting parameter and the matched parameter of FLANN for crossing adjustment SURF characteristic point, to obtain more matching double points, to improve
Registration accuracy；Since reference picture and image to be spliced have certain color difference, can be increased substantially by brightness of image adjustment
Effect after image co-registration.Fabric analysis is carried out using the method for the present invention for the wide-field micro-image of later use to provide effectively
Data.According to different application backgrounds, the present invention is equally applicable to the image mosaic of other related fieldss by modification appropriate.
The foregoing is merely presently preferred embodiments of the present invention, is not intended to limit the invention, it is all in spirit of the invention and
Within principle, any modification, equivalent replacement, improvement and so on be should all be included in the protection scope of the present invention.
Claims (5)
1. a kind of seamless joint method of low coincidence factor micro-image, which is characterized in that method includes the following steps:
Step 1: choosing area-of-interest, that is, ROI with reference to figure and figure to be spliced, ROI is pre-processed；
Step 2: extracting the SURF characteristic point of ROI, characteristic point is slightly matched using FLANN method；
Step 3: the excessive matching pair of space length is rejected according to the deviation range of two images；
Step 4: the discrete distribution of the space length of the remaining matching pair of statistics, to be maximally distributed section corresponding to probability as standard
Range further rejects error hiding pair；If matching returns to abovementioned steps to deficiency, feature extraction parameter and match parameter are modified,
To obtain more matchings pair；
Step 5: according to remainder matching pair, calculating the positional relationship of two images；
Step 6: treating spliced map and carry out brightness adjustment, weighted mean method is taken to merge image；Wherein:
Coincidence factor of the low coincidence factor between adjacent image is between 3%~11%；
The step 3 specifically:
According to the feature of multiple acquired image pattern, image Duplication is between 3% ~ 11%, and the offset of vertical direction is 5%
Within, larger boundary 11% and 5% is taken, obtains corresponding boundary shifts amount, as thresholding, to each pair of in thick matching result
With point, judge whether its space length meets the threshold condition, satisfaction then retains；It is unsatisfactory for, is considered as error hiding and is removed；
The step 4 specifically:
The space length of matching double points is divided into the equal section of multiple step-lengths, by space length discretization；To previous step institute
All matchings pair remained calculate matching to the discrete distribution situation of space length；If secondary maximum probability and maximum probability
Ratio be less than certain threshold value T, then choose section corresponding to maximum probability be critical field；If the ratio is greater than certain threshold value,
And secondary maximum probability is adjacent with section corresponding to maximum probability, then chooses the interlude of the two adjacent intervals as standard model
It encloses；Other situations, then return step 2, change the extracting parameter and the matched parameter of FLANN of SURF characteristic point, to obtain more
More matching double points；Successively judge whether the space length for the matching double points that every a pair remains meets the critical field, it is full
The conduct of sufficient condition is correctly matched to reservation, and the ungratified error hiding that is considered as is to being removed；Judge the matching finally remained
Whether logarithm reaches certain amount, if lazy weight, return step 2 changes the extracting parameter and FLANN of SURF characteristic point
Matched parameter, to obtain more matching double points；Wherein, certain threshold value T value range is 0.4 ~ 0.6.
2. the seamless joint method of low coincidence factor micro-image according to claim 1, which is characterized in that the step 1
Specifically:
Read reference pictureI 0(x, y) and image to be splicedI 1(x, y), it chooses in two width figures and closes on area comprising lap
Domain is as region of interest ROI, using the noise spot in median filtering method removal ROI, wherein filter window is dimensioned to 3
×3；
Using the method for histogram equalization, expand its dynamic range, to enhance the contrast of image.
3. the seamless joint method of low coincidence factor micro-image according to claim 1, which is characterized in that the step 2
Specifically:
The SURF characteristic point in ROI is extracted respectively, and each characteristic point is described with an isometric feature vector, the feature extracted
Point set is denoted as F0 and F1 respectively；
Using FLANN method, for each characteristic point in F0, found out in F1 therewith feature vector apart from the smallest K point,
K=2, if the ratio of minimum range and time small distance is less than certain threshold value, then it is assumed that point corresponding to minimum range is matched；
Otherwise it is assumed that this feature point in F0 is not matched characteristic point in F1；Wherein, certain threshold value initial value is set as
0.5。
4. the seamless joint method of low coincidence factor micro-image according to claim 1, which is characterized in that the step 5
Specifically:
To the matching double points finally remained, according to the spatial position of match point, using the method for average seek reference picture and to
The spatial offset of stitching image determines positional relationship between the two.
5. the seamless joint method of low coincidence factor micro-image according to claim 1, which is characterized in that the step 6
Specifically:
According to the relative positional relationship for the two images that previous step obtains, the real overlapping region of two images is determined；
The RGB color mean value for calculating separately overlapping region between reference picture and image to be spliced obtains each color of reference picture
Each color component the mean value R1, G1, B1 of component mean value R0, G0, B0 and image to be spliced；
The color mean value of image to be spliced subtracts the color mean value of reference picture, obtains color difference Dr, Dg, Db, in this, as
The overall brightness difference of two images；
Each pixel of stitching image is treated, each color component adds Dr, Dg, Db respectively, its luminance level is adjusted to
It is consistent with reference picture；
Each pixel of image overlapping region to be spliced to reference picture and by brightness adjustment, according to its position,
Different weights is set, with the progressive weighted average method gradually gone out, overlapping region is merged, non-overlap area is then directly replicated
Original image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710212589.5A CN107093166B (en) | 2017-04-01 | 2017-04-01 | The seamless joint method of low coincidence factor micro-image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710212589.5A CN107093166B (en) | 2017-04-01 | 2017-04-01 | The seamless joint method of low coincidence factor micro-image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107093166A CN107093166A (en) | 2017-08-25 |
| CN107093166B true CN107093166B (en) | 2019-07-05 |
Family
ID=59648929
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710212589.5A Active CN107093166B (en) | 2017-04-01 | 2017-04-01 | The seamless joint method of low coincidence factor micro-image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107093166B (en) |
Families Citing this family (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108109112B (en) * | 2018-01-16 | 2021-07-20 | 上海同岩土木工程科技股份有限公司 | Tunnel layout graph splicing parameter processing method based on Sift characteristic |
| CN109559275A (en) * | 2018-11-07 | 2019-04-02 | 苏州迈瑞科技有限公司 | A kind of Urine Analyzer MIcrosope image joining method |
| CN110689511B (en) * | 2019-09-19 | 2023-02-03 | 华东师范大学 | Optimization method for large-visual-field image splicing path |
| CN110599404A (en) * | 2019-09-24 | 2019-12-20 | 陕西晟思智能测控有限公司 | Circuit board microscopic image splicing method and device and information data processing terminal |
| CN110738695B (en) * | 2019-10-12 | 2021-08-13 | 哈尔滨工业大学 | Image feature point mismatching and removing method based on local transformation model |
| CN110706191B (en) * | 2019-10-14 | 2022-09-20 | 中国空气动力研究与发展中心超高速空气动力研究所 | High-pressure container detection method based on infrared image splicing fusion algorithm |
| CN111105346A (en) * | 2019-11-08 | 2020-05-05 | 同济大学 | Full-scanning microscopic image splicing method based on peak value search and gray template registration |
| CN111179170B (en) * | 2019-12-18 | 2023-08-08 | 深圳北航新兴产业技术研究院 | Rapid panoramic stitching method for microscopic blood cell images |
| CN113012030A (en) * | 2019-12-20 | 2021-06-22 | 北京金山云网络技术有限公司 | Image splicing method, device and equipment |
| CN111260561A (en) * | 2020-02-18 | 2020-06-09 | 中国科学院光电技术研究所 | Rapid multi-graph splicing method for mask defect detection |
| CN111882484B (en) * | 2020-06-16 | 2024-04-16 | 河北汉光重工有限责任公司 | Servo control method in high-speed seamless image stitching of diving imaging system |
| CN112164001A (en) * | 2020-09-29 | 2021-01-01 | 南京理工大学智能计算成像研究院有限公司 | Digital microscope image rapid splicing and fusing method |
| CN113205457B (en) * | 2021-05-11 | 2023-03-10 | 华中科技大学 | Microscopic image splicing method and system |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104036480A (en) * | 2014-06-20 | 2014-09-10 | 天津大学 | Surf algorithm based quick mismatching point eliminating method |
| CN104851094A (en) * | 2015-05-14 | 2015-08-19 | 西安电子科技大学 | Improved method of RGB-D-based SLAM algorithm |
| CN104966270A (en) * | 2015-06-26 | 2015-10-07 | 浙江大学 | Multi-image stitching method |
| CN105608689A (en) * | 2014-11-20 | 2016-05-25 | 深圳英飞拓科技股份有限公司 | Method and device for eliminating image feature mismatching for panoramic stitching |
| CN105869120A (en) * | 2016-06-16 | 2016-08-17 | 哈尔滨工程大学 | Image stitching real-time performance optimization method |
| CN105957007A (en) * | 2016-05-05 | 2016-09-21 | 电子科技大学 | Image stitching method based on characteristic point plane similarity |
-
2017
- 2017-04-01 CN CN201710212589.5A patent/CN107093166B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104036480A (en) * | 2014-06-20 | 2014-09-10 | 天津大学 | Surf algorithm based quick mismatching point eliminating method |
| CN105608689A (en) * | 2014-11-20 | 2016-05-25 | 深圳英飞拓科技股份有限公司 | Method and device for eliminating image feature mismatching for panoramic stitching |
| CN104851094A (en) * | 2015-05-14 | 2015-08-19 | 西安电子科技大学 | Improved method of RGB-D-based SLAM algorithm |
| CN104966270A (en) * | 2015-06-26 | 2015-10-07 | 浙江大学 | Multi-image stitching method |
| CN105957007A (en) * | 2016-05-05 | 2016-09-21 | 电子科技大学 | Image stitching method based on characteristic point plane similarity |
| CN105869120A (en) * | 2016-06-16 | 2016-08-17 | 哈尔滨工程大学 | Image stitching real-time performance optimization method |
Non-Patent Citations (3)
| Title |
|---|
| Improved SURF Detection combined with Dual FLANN Matching and Clustering Analysis;Li JunFei 等;《Applied Mechanics and Materials Vols》;20140523;全文 |
| 基于SURF特征提取和FLANN搜索的图像匹配搜索;冯亦东 等;《图学学报》;20150831;第36卷(第4期);全文 |
| 基于单目视觉的全景拼接技术研究及安卓平台实现;宋璐;《中国优秀硕士学位论文全文数据库 信息科技辑》;20150731(第7期);全文 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107093166A (en) | 2017-08-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107093166B (en) | The seamless joint method of low coincidence factor micro-image | |
| Du et al. | Image segmentation-based multi-focus image fusion through multi-scale convolutional neural network | |
| CN106650794B (en) | A kind of image bloom removing method influenced by body surface high light reflectivity and system | |
| US20130223760A1 (en) | Microscopic image fusion method based on region growing | |
| CN109714519B (en) | Method and system for automatically adjusting image frame | |
| CN108399611A (en) | Multi-focus image fusing method based on gradient regularisation | |
| Xiao et al. | Multi-focus image fusion based on depth extraction with inhomogeneous diffusion equation | |
| CN104182952B (en) | Multi-focus sequence image fusion method | |
| Zadeh et al. | Recognition and classification of the cancer cells by using image processing and labview | |
| CN104992447A (en) | Automatic image detection method for moving microorganisms in sewage | |
| CN111105346A (en) | Full-scanning microscopic image splicing method based on peak value search and gray template registration | |
| CN110705634B (en) | Heel model identification method and device and storage medium | |
| CN107909588A (en) | Partition system under MRI cortex based on three-dimensional full convolutional neural networks | |
| CN104899834A (en) | Blurred image recognition method and apparatus based on SIFT algorithm | |
| CN110310291A (en) | A kind of rice blast hierarchy system and its method | |
| CN112087970A (en) | Information processing apparatus, information processing method, and computer program | |
| CN108734145A (en) | A kind of face identification method based on degree adaptive face characterization model | |
| CN112037185A (en) | Chromosome split phase image screening method and device and terminal equipment | |
| CN110263778A (en) | A kind of meter register method and device based on image recognition | |
| CN100423021C (en) | Method and device for segmentation low depth image | |
| CN113763300A (en) | Multi-focus image fusion method combining depth context and convolution condition random field | |
| Xiao et al. | Hematoxylin and eosin (H&E) stained liver portal area segmentation using multi-scale receptive field convolutional neural network | |
| CN116596895A (en) | Substation equipment image defect identification method and system | |
| Ho et al. | AF-Net: A convolutional neural network approach to phase detection autofocus | |
| CN113436120A (en) | Image fuzzy value identification method and device |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |