CN112837235B - Neighborhood-based adaptive spatial filtering method - Google Patents
Neighborhood-based adaptive spatial filtering method Download PDFInfo
- Publication number
- CN112837235B CN112837235B CN202110104104.7A CN202110104104A CN112837235B CN 112837235 B CN112837235 B CN 112837235B CN 202110104104 A CN202110104104 A CN 202110104104A CN 112837235 B CN112837235 B CN 112837235B
- Authority
- CN
- China
- Prior art keywords
- pixel
- point
- adaptive
- filtering
- neighborhood
- 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
- 238000001914 filtration Methods 0.000 title claims abstract description 63
- 238000000034 method Methods 0.000 title claims abstract description 37
- 230000003044 adaptive effect Effects 0.000 title claims abstract description 24
- 238000012163 sequencing technique Methods 0.000 claims abstract description 5
- 239000010453 quartz Substances 0.000 claims description 4
- VYPSYNLAJGMNEJ-UHFFFAOYSA-N silicon dioxide Inorganic materials O=[Si]=O VYPSYNLAJGMNEJ-UHFFFAOYSA-N 0.000 claims description 4
- 238000001514 detection method Methods 0.000 description 16
- 238000010586 diagram Methods 0.000 description 15
- 230000002146 bilateral effect Effects 0.000 description 5
- 235000002566 Capsicum Nutrition 0.000 description 4
- 239000006002 Pepper Substances 0.000 description 4
- 241000722363 Piper Species 0.000 description 4
- 235000016761 Piper aduncum Nutrition 0.000 description 4
- 235000017804 Piper guineense Nutrition 0.000 description 4
- 235000008184 Piper nigrum Nutrition 0.000 description 4
- 150000003839 salts Chemical class 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- G06T5/70—
-
- 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
- G06T2207/10032—Satellite or aerial image; Remote sensing
-
- 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
- G06T2207/20024—Filtering details
- G06T2207/20028—Bilateral filtering
Landscapes
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a neighborhood-based adaptive spatial filtering method, which comprises the following steps: step 1: initializing each pixel point of an input image, and acquiring a pixel value of a corresponding window; step 2: sequencing from small to large to obtain upper quartile pixel points and lower quartile pixel points; and step 3: performing adaptive neighborhood expansion by taking the quartile pixel points and the lower quartile pixel points as constraint conditions, performing adaptive filtering on the obtained adaptive region serving as neighborhood information to obtain a filtered image, and calculating the coordinates of all the pixel points in the adaptive region and the variance of the seed pixel values of the pixel points; and 4, step 4: obtaining the variance of the filtered image and the seed pixel points, and judging whether the conditions are satisfied; and 5: and (5) repeating the step (3) - (step (4)), iterating and calculating until all the pixel points meet the judgment condition, and taking the obtained last filtered image as a final filtered image. The method of the invention keeps the detail information of the image edge and has high definition.
Description
Technical Field
The invention belongs to the technical field of neighborhood change detection of remote sensing images, and relates to a neighborhood-based adaptive spatial filtering method.
Background
With the rapid development of remote sensing satellite and aerospace remote sensing technologies, the spatial resolution and the time resolution of high-resolution images are greatly improved, and through analysis and rapid processing of remote sensing image data, earth surface change information can be effectively acquired, data support is provided for exploring earth surface change conditions, and important scientific basis is provided for aspects such as city change, precision agriculture and geological disasters. On one hand, the capability of acquiring the ground feature information is enhanced along with the improvement of the image resolution, but on the other hand, the micro-noise of new and old images in different periods is found to be more, and the spectra of different ground features are mutually overlapped, so that the variance among the ground features is increased.
In contrast, the traditional filtering method can filter salt and pepper noise, and simultaneously mix different ground objects, so that the boundary is blurred, and the reliability and the accuracy of a change detection result are greatly reduced.
Disclosure of Invention
The invention aims to provide a neighborhood-based adaptive spatial filtering method, which solves the problems that different ground objects are mixed and the boundary is fuzzy while salt and pepper noise is filtered by the filtering method in the prior art, and the reliability and the accuracy of a change detection result are greatly reduced.
The technical scheme adopted by the invention is that a neighborhood-based adaptive spatial filtering method is implemented according to the following steps:
step 1: initializing each pixel point of an input image, and acquiring a pixel value of a corresponding window;
step 2: counting each obtained pixel value, and sequencing from small to large to obtain an upper quartile pixel point Up _ quartz _ pixel and a lower quartile pixel point Low _ quartz _ pixel;
and step 3: aiming at each Pixel point as a seed Pixel point Pixel, self-adaptive neighborhood expansion is carried out by taking the upper quartile Pixel point and the lower quartile Pixel point as constraint conditions, if the value of the Pixel meets the condition that Low _ quality _ Pixel is not less than Pixel is not more than Up _ quality _ Pixel, all the Pixel points meeting the condition are marked as self-adaptive areas, the obtained self-adaptive areas are taken as neighborhood information to carry out self-adaptive filtering to obtain a filtering image, and the coordinates of all the Pixel points in the self-adaptive areas and the variance Std of the seed Pixel value Pixel thereof are calculated n ；
And 4, step 4: taking each pixel point in the filtered image obtained in the step 3 as a seed pixel point, and taking the coordinate point recorded in the step 3 as an extension point coordinate to perform self-adaptive filtering to obtain a variance Std of the filtered image and the seed pixel point (n+1) Judging | | | Std (n+1) -Std n Whether | | ≦ epsilon or not, wherein N ∈ N + If yes, the seed point does not participate in the next filtering;
if the seed point is not established, the seed point continues to participate in the filtering next time until the Std | | (n+1) -Std n If | | ≦ epsilon;
and 5: and (4) repeating the operation of the step (3) - (4) on the filtered image obtained in the step (4), performing iterative calculation until all pixel points meet the judgment condition, stopping iteration, and taking the obtained last filtered image as a final filtered image.
The method has the advantages that the noise can be effectively reduced, meanwhile, the detail information such as the image edge and the like is kept, the definition among different types of ground objects is kept, the intra-class homogeneity is improved, and the automation degree is obviously improved.
Drawings
FIG. 1 is a schematic flow diagram of the process of the present invention;
FIG. 2 is a graph of the filtering effect of an embodiment based on a Quickbird satellite image, wherein FIG. 2a is an RGF edge preserving filter graph; FIG. 2b is a graph of Guided filtering; FIG. 2c is a diagram of Bilateral filtering; FIG. 2d is a diagram of RFs filtering; FIG. 2e is a graph of Mean filtering; FIG. 2f is a diagram of adaptive spatial filtering for the method of the present invention;
FIG. 3 is a graph of the results of post-filter change detection using different methods, and FIG. 3a is a T1 time period image; FIG. 3b is a T2 epoch image; FIG. 3c is a reference true value graph; FIG. 3d is a RGF filter graph; FIG. 3e is a Guided filtering filter graph; FIG. 3f is a diagram of Bilateral filtering; FIG. 3g is a diagram of RFs filtering; FIG. 3h is a graph of Mean filtering; fig. 3i is a diagram of the adaptive spatial filtering of the method of the present invention.
Detailed Description
The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.
Referring to fig. 1, the adaptive spatial filtering method of the present invention is implemented according to the following steps:
step 1: initializing each pixel point of an input image, and acquiring a corresponding window pixel value by using a window with the size of 3 multiplied by 3;
step 2: counting each pixel value obtained by a 3 multiplied by 3 window, and sequencing the pixel values from small to large (because an upper quartile point is smaller than a lower quartile point, the quartile naming and conclusion have been proved, if the pixel values are from large to small, the opposite is true (the upper quartile point is larger than the lower quartile point), and the constraint conditions are contradicted later), and obtaining an upper quartile pixel point Up _ pixel and a lower quartile pixel point Low _ pixel;
and 3, step 3: aiming at each Pixel point as a seed Pixel point Pixel, self-adaptive neighborhood expansion is carried out by taking an upper quartile Pixel point and a lower quartile Pixel point as constraint conditions, if the value of the Pixel meets the condition of Low _ quality _ Pixel being not less than Pixel being not more than Up _ quality _ Pixel, all the Pixel points meeting the condition are marked as self-adaptive areas, the obtained self-adaptive areas are taken as neighborhood information to carry out self-adaptive filtering to obtain a filtering image, and the coordinates of all the Pixel points in the self-adaptive areas and the variance Std of the sub-Pixel value Pixel thereof are calculated n ；
And 4, step 4: taking each pixel point in the filtered image obtained in the step 3 as a seed pixel point, and taking the coordinate point recorded in the step 3 as an extension point coordinate to perform self-adaptive filtering to obtain a variance Std of the filtered image and the seed pixel point (n+1) Judging | | | Std (n+1) -Std n Whether | ≦ ε (i.e., std) n And Std (n+1) Corresponding to the same seed point), wherein N is belonged to N + Epsilon is any decimal number, preferably epsilon =0.005,
if yes, the seed point does not participate in the next filtering;
if the seed point is not established, the seed point continues to participate in the filtering next time until the Std | | (n+1) -Std n If | | < epsilon;
and 5: and (4) repeating the operation of the step (3) - (4) on the filtered image obtained in the step (4), performing iterative calculation until all pixel points meet the judgment condition, stopping iteration, and taking the obtained last filtered image as a final filtered image.
Example (b):
the detailed steps of the invention are as follows:
description of image data: the JN data is raw image data acquired through a QuickBird satellite, and includes spectral ranges of a blue band (0.45-0.52 m), a green band (0.52-0.59 m), a red band (0.63-0.69 m), and a near-infrared band (0.77-0.89 m), and the raw image includes a plurality of types of feature scenes.
And 2, counting each pixel value obtained by a 3 multiplied by 3 window, and sequencing (from small to large) to obtain an upper quartile pixel point Up _ quartile _ pixel and a lower quartile pixel point Low _ quartile _ pixel.
Step 3, aiming at each Pixel point as a seed Pixel point Pixel, performing adaptive neighborhood point expansion by taking an upper quartile Pixel point and a lower quartile Pixel point as constraint conditions (Low _ quality _ Pixel is less than or equal to Pixel is less than or equal to Up _ quality _ Pixel), marking all Pixel points meeting the conditions as adaptive areas, performing adaptive filtering by taking the obtained adaptive areas as neighborhood information to obtain a filtered image, and simultaneously recording the coordinates of the obtained adaptive Pixel points and the variance Std of the seed Pixel points n 。
Step 4, taking each pixel point in the filtered image obtained in the step 3 as a seed pixel point, and taking the coordinate point recorded in the step 3 as an extension point coordinate to perform self-adaptive filtering to obtain the variance Std of the filtered image and the seed pixel point (n+1) Judging | | | Std (n+1) -Std n Whether | ≦ ε (Std) n And Std (n+1) Corresponding to the same seed point), where N ∈ N + If epsilon is any decimal, epsilon =0.005 is selected, and if epsilon is satisfied, the seed point does not participate in the next filtering; if not, continuing to participate in the filtering next time until the Std | | (n+1) -Std n If | | ≦ epsilon.
And 5, repeating the operation of the step 3-the step-4 on the filtered image obtained in the step 4, carrying out iterative calculation until all pixel points meet the conditions, stopping iteration, and taking the obtained last filtered image as a final filtered image.
And (3) experimental verification:
comparing the filtering method of the present invention with five other existing filtering methods, wherein fig. 2a edge-preserving filtering graph; FIG. 2b is a guided filtering diagram; FIG. 2c is a bilateral filter graph; FIG. 2d RFs filter diagram; FIG. 2e mean filtering graph; fig. 2f is a diagram of adaptive spatial filtering, which is a filtering method proposed by the present invention, and compared with the five existing methods, the method of the present invention can reduce salt and pepper noise in the high-resolution remote sensing image detection process, simultaneously retain the boundary characteristics of the region, reduce the variance between the same ground objects, and improve the homogeneity.
Comparing the change detection results of the filtering method with the other five existing filtering methods to obtain a result graph of change detection after filtering under different method conditions, wherein the graph of fig. 3a is an image in the period T1; FIG. 3b is a T2 epoch image; FIG. 3c is a reference true value graph; FIG. 3d is a RGF [1] filter graph; FIG. 3e is a Guided filtering [2] filter diagram; FIG. 3f is a diagram of Bilateral filtering (3); FIG. 3g is a diagram of RFs [4] filtering; FIG. 3h is a Mean filtering [5] filter plot (window parameter 3 × 3); fig. 3i is a diagram of the adaptive spatial filtering of the method of the present invention. Compared with the prior filtering method, the method can generate better results in the field of change detection, can effectively improve the change detection effect and improve the change detection precision.
The obtained change detection results were compared with the truth map to obtain quantitative comparison results, as shown in table 1. Table 1 shows the accuracy comparison table in the change detection of the method of the present invention and different methods, and it is obvious that the method of the present invention can significantly improve the accuracy when the high resolution remote sensing image is applied to the table coverage change detection, compared with the similar methods.
TABLE 1 accuracy quantitative comparison based on Quickbird satellite images
| OA | Kappa | AA | FA | MA | TA | Completeness | Correctness | |
| Edge preserving filter map | 90.18 | 0.899 | 92.54 | 11.02 | 3.89 | 9.82 | 96.11 | 63.92 |
| Guided filter graph | 90.24 | 0.9 | 92.75 | 11.04 | 3.465 | 9.76 | 96.54 | 63.98 |
| Bilateral filter graph | 90.03 | 0.898 | 92.4 | 11.17 | 4.036 | 9.965 | 95.96 | 63.57 |
| RFs filter diagram | 90.02 | 0.898 | 92.75 | 11.37 | 3.136 | 9.976 | 96.86 | 63.38 |
| Mean filter graph | 84.76 | 0.843 | 89.67 | 17.74 | 2.925 | 15.24 | 97.07 | 52.64 |
| Adaptive spatial filtering | 94.53 | 0.944 | 94.13 | 5.271 | 6.468 | 5.473 | 93.53 | 78.28 |
In conclusion, the method of the invention can reduce the salt and pepper noise in the detection process of two high-resolution remote sensing images in different periods, and can also retain the boundary characteristics of the change area, thereby improving the change detection precision of the high-resolution remote sensing images and effectively improving the change detection effect.
Reference documents:
1.Zhang,Q.；Shen,X.；Xu,L.；Jia,J.In Rolling guidance filter,European conference on computer vision,2014；Springer:pp 815-830.
2.He,K.；Sun,J.；Tang,X.In Guided image filtering,European conference on computer vision,2010；Springer:pp 1-14.
3.Xing,Q.；Chen,C.；Li,Z.Progressive path tracing with bilateral-filtering-based denoising.Multimedia Tools and Applications 2020,1-16.
4.Kang,X.；Li,S.；Benediktsson,J.A.Feature extraction of hyperspectral images with image fusion and recursive filtering.IEEE Transactions on Geoscience and Remote Sensing 2013,52,3742-3752.
5.Rakshit,S.；Ghosh,A.；Shankar,B.U.Fast mean filtering technique(fmft).Pattern Recognition 2007,40,890-897.
Claims (3)
1. a neighborhood-based adaptive spatial filtering method is characterized by comprising the following steps:
step 1: initializing each pixel point of an input image, and acquiring a pixel value of a corresponding window;
step 2: counting each obtained pixel value, and sequencing the pixel values from small to large to obtain an upper quartile pixel point Up _ quartz _ pixel and a lower quartile pixel point Low _ quartz _ pixel;
and step 3: aiming at each Pixel point as a seed Pixel point Pixel, performing self-adaptive neighborhood expansion by taking the upper quartile Pixel point and the lower quartile Pixel point as constraint conditions, marking all Pixel points meeting the conditions as self-adaptive areas if the value of the Pixel meets the condition that Low _ quality _ Pixel is not more than Pixel is not more than Up _ quality _ Pixel, performing self-adaptive filtering by taking the obtained self-adaptive areas as neighborhood information to obtain a filtered image, and calculating the self-adaptive areasThere is coordinate of Pixel point and variance Std of Pixel value Pixel n ；
And 4, step 4: taking each pixel point in the filtered image obtained in the step 3 as a seed pixel point, and taking the coordinate point recorded in the step 3 as an extension point coordinate to perform self-adaptive filtering to obtain a variance Std of the filtered image and the seed pixel point (n+1) Judging | | | Std (n+1) -Std n Whether | ≦ epsilon or not, wherein N ∈ N + If yes, the seed point does not participate in the next filtering;
if the seed point is not established, the seed point continues to participate in the filtering next time until the Std | | (n+1) -Std n If | | < epsilon;
and 5: and (5) repeating the operation of the step (3) - (step (4)) on the filtered image obtained in the step (4), performing iterative computation until all pixel points meet the judgment condition, stopping iteration, and taking the obtained last filtered image as a final filtered image.
2. The neighborhood-based adaptive spatial filtering method according to claim 1, wherein: in step 1, the pixel values of the corresponding window are obtained by using a window with a size of 3 × 3.
3. The neighborhood-based adaptive spatial filtering method according to claim 1, wherein: in step 4, epsilon =0.005.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110104104.7A CN112837235B (en) | 2021-01-26 | 2021-01-26 | Neighborhood-based adaptive spatial filtering method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110104104.7A CN112837235B (en) | 2021-01-26 | 2021-01-26 | Neighborhood-based adaptive spatial filtering method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112837235A CN112837235A (en) | 2021-05-25 |
| CN112837235B true CN112837235B (en) | 2022-12-13 |
Family
ID=75931827
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110104104.7A Active CN112837235B (en) | 2021-01-26 | 2021-01-26 | Neighborhood-based adaptive spatial filtering method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112837235B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106105201A (en) * | 2014-03-14 | 2016-11-09 | 高通股份有限公司 | Use the de-blocking filter of pixel distance |
| CN106910169A (en) * | 2017-01-26 | 2017-06-30 | 苏州大学 | A kind of image salt-pepper noise minimizing technology for preventing edge blurry |
| CN108389162A (en) * | 2018-01-09 | 2018-08-10 | 浙江大学 | A kind of image border holding filtering method based on adaptive neighborhood shape |
| CN108765336A (en) * | 2018-05-25 | 2018-11-06 | 长安大学 | Image defogging method based on dark bright primary colors priori with auto-adaptive parameter optimization |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9576345B2 (en) * | 2015-02-24 | 2017-02-21 | Siemens Healthcare Gmbh | Simultaneous edge enhancement and non-uniform noise removal using refined adaptive filtering |
| WO2016204531A1 (en) * | 2015-06-16 | 2016-12-22 | 엘지전자(주) | Method and device for performing adaptive filtering according to block boundary |
-
2021
- 2021-01-26 CN CN202110104104.7A patent/CN112837235B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106105201A (en) * | 2014-03-14 | 2016-11-09 | 高通股份有限公司 | Use the de-blocking filter of pixel distance |
| CN106910169A (en) * | 2017-01-26 | 2017-06-30 | 苏州大学 | A kind of image salt-pepper noise minimizing technology for preventing edge blurry |
| CN108389162A (en) * | 2018-01-09 | 2018-08-10 | 浙江大学 | A kind of image border holding filtering method based on adaptive neighborhood shape |
| CN108765336A (en) * | 2018-05-25 | 2018-11-06 | 长安大学 | Image defogging method based on dark bright primary colors priori with auto-adaptive parameter optimization |
Non-Patent Citations (3)
| Title |
|---|
| "Improved adaptive median filters using nearest 4-neighbors for restoration of images corrupted with fixed-valued impulse noise";S.Samsad Beagum 等;《IEEE》;20160321;全文 * |
| 一种自适应双阈值中值滤波方法;宗永胜等;《小型微型计算机系统》;20170715(第07期);220-225 * |
| 基于噪声点多级检测的自适应中值滤波算法;朱威等;《光电工程》;20131015(第10期);67-73 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112837235A (en) | 2021-05-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Dubey et al. | FCM clustering algorithms for segmentation of brain MR images | |
| CN108765465B (en) | Unsupervised SAR image change detection method | |
| CN109102469B (en) | Remote sensing image panchromatic sharpening method based on convolutional neural network | |
| CN109146890B (en) | Abnormal target detection method of hyperspectral image based on filter | |
| CN111369605B (en) | Infrared and visible light image registration method and system based on edge features | |
| CN112884791B (en) | Method for constructing large-scale remote sensing image semantic segmentation model training sample set | |
| CN107169962B (en) | Gray level image fast segmentation method based on space density constraint kernel fuzzy clustering | |
| CN106296763B (en) | A kind of metal material Industry CT Image Quality method for quickly correcting | |
| CN111369457A (en) | Remote sensing image denoising method for sparse discrimination tensor robustness PCA | |
| CN109671029A (en) | Image denoising algorithm based on gamma norm minimum | |
| CN111091111A (en) | Vehicle bottom dangerous target identification method | |
| Yamanappa et al. | Non-local means image denoising using Shapiro-Wilk similarity measure | |
| CN110458773B (en) | Anisotropic diffusion noise processing method based on edge enhancement operator | |
| CN115797381A (en) | Heterogeneous remote sensing image registration method based on geographic blocking and hierarchical feature matching | |
| CN113421210B (en) | Surface point Yun Chong construction method based on binocular stereoscopic vision | |
| Rao et al. | Robust optical flow estimation via edge preserving filtering | |
| CN112837235B (en) | Neighborhood-based adaptive spatial filtering method | |
| CN113344810A (en) | Image enhancement method based on dynamic data distribution | |
| CN109635809A (en) | A kind of superpixel segmentation method towards vision degraded image | |
| Tian et al. | A novel diffusion system for impulse noise removal based on a robust diffusion tensor | |
| CN112802042A (en) | Two-dimensional core digital image particle segmentation method and device and readable storage medium | |
| CN113436123B (en) | High-resolution SAR and low-resolution multispectral image fusion method based on cloud removal-resolution improvement cooperation | |
| CN115511928A (en) | Matching method of multispectral image | |
| CN115841618A (en) | Remote sensing image coastline extraction method based on clustering and edge detection | |
| CN114926360A (en) | Image noise reduction processing working method based on noise estimation |
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 |