CN116309034A - Optimal spelling line acquisition method for ultra-large file remote sensing image - Google Patents
Optimal spelling line acquisition method for ultra-large file remote sensing image Download PDFInfo
- Publication number
- CN116309034A CN116309034A CN202211104116.0A CN202211104116A CN116309034A CN 116309034 A CN116309034 A CN 116309034A CN 202211104116 A CN202211104116 A CN 202211104116A CN 116309034 A CN116309034 A CN 116309034A
- Authority
- CN
- China
- Prior art keywords
- image
- remote sensing
- line
- optimal
- oversized
- 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
- 238000000034 method Methods 0.000 title claims abstract description 33
- 239000000872 buffer Substances 0.000 claims abstract description 40
- 238000005070 sampling Methods 0.000 claims abstract description 8
- 230000003595 spectral effect Effects 0.000 claims description 37
- 238000012952 Resampling Methods 0.000 claims description 3
- 238000001228 spectrum Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 8
- 238000004364 calculation method Methods 0.000 description 5
- 238000005457 optimization Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004040 coloring Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000000265 homogenisation Methods 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 230000002035 prolonged effect Effects 0.000 description 1
- 230000011218 segmentation Effects 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
- G06T3/4007—Scaling of whole images or parts thereof, e.g. expanding or contracting based on interpolation, e.g. bilinear interpolation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/12—Edge-based segmentation
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Processing (AREA)
Abstract
The invention provides an optimal splicing line acquisition method of an ultra-large file remote sensing image, which comprises the following steps: acquiring two ultra-large file remote sensing images with geographic coordinate references, performing equal resolution downsampling treatment on the overlapped area of the two ultra-large file remote sensing images, and establishing an undirected graph for pixel points after the resolution is reduced; searching the minimum cut of the undirected graph by using a graph cut algorithm, and obtaining a rough spelling line; generating a buffer based on the rough stitching line; obtaining a banded overlapping region of two scenery images obtained by mapping the buffer region in the original oversized file remote sensing image; establishing an undirected graph for pixel points in the banded overlapping region; and obtaining the optimal splicing line of the banded overlapping region by applying a graph cut algorithm. According to the invention, through the resolution sampling of the ultra-large file remote sensing image, the number of pixel points to be traversed in the determining process of the stitching line is reduced, so that the number of nodes and edges to be calculated in a graph cutting algorithm is met, and the nodes and edges are ensured to be global optimal solutions, thereby obtaining the optimal stitching line of the ultra-large file remote sensing image.
Description
Technical Field
The invention relates to the field of splicing of remote sensing images, in particular to an optimal splicing line acquisition method of an ultra-large file remote sensing image.
Background
Image stitching is a technique of stitching several images (possibly acquired at different times and from different sensors) with overlapping portions into a seamless panoramic image or high resolution image. In the application of remote sensing images, in order to obtain a larger field of view, and better uniformly process, analyze, study and interpret remote sensing image information, two or more obtained remote sensing images with overlapping areas need to be spliced into an image. The big data remote sensing image mainly refers to a satellite remote sensing image, and the image data size of the big data remote sensing image gradually becomes larger along with the improvement of the resolution of the satellite remote sensing image. The Graph Cut algorithm (Graph Cut) is a very useful and popular energy optimization algorithm, and is widely applied to front background segmentation (Image segmentation), stereoscopic vision (stereo vision), matting (Image matting) and the like in the field of computer vision. Because of the energy-optimized nature of the graph cut algorithm, it can be applied to find the optimal stitching line between images by assigning values to the image edges, so that the spectral difference between the pixels on both sides of the image stitching line is minimal. However, the current method for searching the optimal spelling line between two images by establishing an undirected graph and applying a graph cutting algorithm is limited to the maximum node number and the maximum edge number which can be accommodated by the graph cutting algorithm, so that the universality of the method for remote sensing images is greatly limited. For example, the number of pixels in the overlapping area of two big data remote sensing images is 5000×5000, and the maximum capacity that can be calculated by the graph cut algorithm is 1000×1000, so that the situation that the graph cut algorithm cannot be applied can occur. In addition, the simple direct application of the graph cut algorithm can also lead to the addition of a plurality of invalid points to calculation, so that the calculation time is greatly prolonged.
The traditional image stitching optimization algorithm is only suitable for the condition of small data images, and when the traditional image stitching optimization algorithm is applied to ultra-large data remote sensing images, the condition that the data volume is too large to calculate usually occurs, and the stitching between the remote sensing images with increasingly high resolution and increasingly high image data volume cannot be ensured.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides an optimal spelling line acquisition method of an ultra-large file remote sensing image. Aiming at the defects that calculation cannot be performed or time is excessively long due to excessively large data volume in the process of finding the two-scene oversized file remote sensing image stitching line, the method is used for finding the global optimal stitching line of the two-scene remote sensing image by nesting and applying a graph cutting algorithm. It should be pointed out that the invention is characterized in that nested searching splice lines, unlike the processing of one-time searching or block searching splice lines of common splice lines, the invention downsamples the original image to obtain a reduced image, finds a rough splice line, then makes a buffer zone, searches finer splice lines in the buffer zone, and finally performs splicing and color homogenization based on the splice lines to obtain a seamless spliced big data remote sensing image.
In order to achieve the above purpose, the specific technical scheme of the invention is as follows:
an optimal spelling line acquisition method of an ultra-large file remote sensing image comprises the following steps:
step one: acquiring two ultra-large file remote sensing images with geographic coordinate references, and performing equal resolution downsampling on the overlapped area of the two ultra-large file remote sensing images to acquire a rough splicing line;
the first step is realized by the following substeps:
(1.1) calculating the geographic overlapping area of the remote sensing images of the two-scene oversized file;
(1.2) calculating the quantity of pixels in the geographic overlapping area of the remote sensing images of the two-scene oversized file and the resampling resolution of image stitching;
(1.3) performing equal resolution downsampling treatment on the overlapping area of the remote sensing images of the two-scene oversized file;
(1.4) establishing an undirected graph for pixel points of an overlapping area of the remote sensing images of the two-scene oversized file after the downsampling treatment;
(1.5) searching the minimum cut of the undirected graph obtained in the step (1.4) by using a graph cut algorithm to obtain a rough split joint line after resolution reduction;
step two: generating a buffer area based on the rough spelling line obtained in the first step, wherein the buffer area forms a banded overlapping area in the remote sensing image of the original two-scene oversized file, and the global optimal spelling line is obtained in the banded overlapping area;
the second step is specifically realized by the following substeps:
(2.1) generating a buffer based on the rough stitching line obtained in step one;
(2.2) obtaining a banded overlapping region of the two-scene oversized file remote sensing image formed in the original oversized file remote sensing image by the buffer area generated in the step (2.1);
(2.3) establishing an undirected graph for pixel points in the banded overlapping region;
and (2.4) searching the minimum cut of the undirected graph obtained in the step (2.3) by using a graph cut algorithm, and obtaining the optimal splicing line of the banded overlapping region, namely the global optimal splicing line.
Further, the step (1.3) includes:
and (3) selecting a sampling coefficient n, wherein n is more than 1, downsampling the two-scene oversized file remote sensing image A, B in the first step by using a bilinear interpolation method to obtain two reduced images NA and NB which are reduced by the same proportion compared with A, B, and taking the two reduced images NA and NB as images for searching for rough spelling lines.
Further, the step (1.4) includes:
determining the range of the overlapping area on the contracted images NA and NB, and establishing an undirected graph G with weight information on each side based on the pixel points in the overlapping area 1 (V, E) the undirected graph G 1 (V, E) includes node V { C (x) 1 ，y 1 )，C(x 1 ，y 2 )，…，C(x 1 ，y t )，…，C(x k ，y t ) -and edges E { R (L), S (L), B (L) }; wherein C (x) 1 ，y 1 ) Is expressed in the position (x 1 ，y 1 ) Spectral values at C (x 1 ，y 2 ) Represented at (x 1 ，y 2 ) Spectral values at C (x 1 ，y t ) Represented at (x 1 ，y t ) Spectral values at C (x k ，y t ) Represented at (x k ，y t ) Spectral values at; r (L) represents edges between the same rows of internal nodes, S (L) represents edges between different rows of internal nodes, and B (L) represents the outermost edge of the frame;
undirected graph G 1 The assignment rules for each node and edge in (V, E) are as follows:
edges between the same rows:
R(x，y)＝|C NA (x 1 ，y 1 )-C NB (x 1 ，y 1 )|+|C NA (x 2 ，y 1 )-C NB (x 2 ，y 1 ) Side between different rows, | (1):
S(x，y)＝|C NA (x 1 ，y 1 )-C NB (x 1 ，y 1 )|+|C NA (x 1 ，y 2 )-C NB (x 1 ，y 2 )| (2)
the outermost edges of the frame:
in the formula, the value of the node is the spectrum value of the pixel point corresponding to the remote sensing image of the original oversized file: c (C) NA (x 1 ，y 1 ) For image NA at position (x 1 ，y 1 ) Spectral value at C NA (x 2 ，y 1 ) For image NA at position (x 2 ，y 1 ) Spectral value at C NA (x 1 ，y 2 ) For image NA at position (x 1 ，y 2 ) Spectral value at C NA (x p ，y q ) For image NA at position (x p ，y q ) Spectral values at; c (C) NB (x 1 ，y 1 ) For image NB in position (x 1 ，y 1 ) Spectral value at C NB (x 2 ，y 1 ) For image NB in position (x 2 ，y 1 ) Spectral value at C NB (x 1 ，y 2 ) For image NB in position (x 1 ，y 2 ) Spectral value at C NB (x p ，y q ) For image NB in position (x p ，y q ) Spectral values at.
Further, the step (2.1) includes:
after the rough splicing line is obtained, selecting a buffer area with the radius r, taking buffer areas towards two sides along the normal direction of the rough splicing line, and up-sampling and expanding the buffer areas to the original image size to obtain an image P.
Further, the step (2.2) includes:
mapping the positions of the buffer areas in the image P onto the original image A, B one by one, obtaining a banded overlapping region on the image A, B, and searching for an optimal splicing line in the banded overlapping region;
further, the step (2.3) includes:
establishing an undirected graph G based on pixel points in the banded overlapping region 2 (V, E) wherein the assignment rules for each node and edge are the same as the assignment rules for each node and edge in step (2.4). The beneficial effects of the invention are as follows:
(1) According to the invention, the global optimal stitching line of the two-scene remote sensing image is found by adopting the nested application graph cut algorithm, so that the problem that the traditional stitching line finding method cannot calculate or consumes too much time due to too large data volume in the stitching line finding process is avoided, and compared with the traditional stitching line finding method, the number of nodes for image calculation is greatly reduced, the calculated amount is further reduced, and the calculation efficiency of stitching line finding is improved.
(2) The invention is suitable for finding the splice lines of various sensors and has universal applicability to different sensors.
Drawings
FIG. 1 is a flow chart diagram of an optimal patch line acquisition method for ultra-large file remote sensing images of the present invention;
FIG. 2 is a schematic diagram of a downsampled two-view image overlap region in an embodiment of the present invention;
FIG. 3 is a schematic diagram of establishing multiple buffers based on the obtained rough stitching lines in one embodiment of the present invention;
FIG. 4 (a) is the position of the resulting coarse tile in the overlap region within the buffer after upsampling in one embodiment of the present invention;
FIG. 4 (b) is a position of the resulting globally optimal patch cord within the buffer within a banded overlap region in one embodiment of the present invention;
FIG. 5 (a) is a schematic panoramic view of the result after stitching and color evening in one embodiment of the invention;
fig. 5 (b) is a schematic diagram after selecting the splicing result in fig. 5 (a) to be partially enlarged.
Detailed Description
The objects and effects of the present invention will become more apparent from the following detailed description of the preferred embodiments and the accompanying drawings, in which the present invention is further described in detail. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
As shown in fig. 1, the method for obtaining the optimal spelling line of the remote sensing image of the oversized file according to the invention comprises the following steps:
step one: and acquiring two ultra-large file remote sensing images (Landsat 8) with geographic coordinate references, and performing equal resolution downsampling on the overlapping areas of the two images to acquire a first rough stitching line.
The first step is realized by the following substeps:
1.1. calculating the geographic overlapping area of the two-scene oversized file remote sensing images;
1.2. calculating the number of pixels in the geographic overlapping area of the two-scene images to obtain 9, 048 and 576 pixels in the overlapping area; calculating resampling resolution of image stitching;
1.3. and performing equal resolution downsampling processing on the overlapping area of the two images. The method comprises the following specific steps:
and selecting a pixel block with a sampling coefficient of 10, namely 10 x 10 on the original image as one pixel point in the reduced image. The original image A, B is downsampled by a bilinear interpolation method (Bilinear Interpolation) to obtain two reduced images NA and NB reduced by the same proportion as compared with the original image A, B, so that the number of pixels in the geographic overlapping region in the reduced image is reduced to 62697, and the overlapping region is used as a range for finding a rough stitching line.
Fig. 2 is a schematic diagram of the embodiment after downsampling the overlapping area of the two images. In the figure, the black region is the overlapping region, and is a position in the downsampled image NA.
1.4. And establishing an undirected image for pixel points in the overlapping area of the two images after the downsampling treatment. The method comprises the following specific steps:
determining the range of an overlapping area on a contracted image, and establishing an undirected graph G with weight information at each side based on pixel points in the overlapping area 1 (V, E) comprising node V { C (x) 1 ，y 1 )，C(x 1 ，y 2 )，…，C(x 1 ，y t )，…，C(x k ，y t ) -and edges E { R (L), S (L), B (L) }; wherein C (x) 1 ，y 1 ) Is expressed in the position (x 1 ，y 1 ) Spectral values at C (x 1 ，y 2 ) Represented at (x 1 ，y 2 ) Spectral values at C (x 1 ，y t ) Represented at (x 1 ，y t ) Spectral values at C (x k ，y t ) Represented at (x k ，y t ) Spectral values at; r (L) represents the edges between the same rows of the internal nodes, S (L) represents the edges between different rows of the internal nodes, and B (L) represents the outermost edges of the frame for performing the graph cut algorithm.
Undirected graph G 1 The assignment rules for each node and edge in (V, E) are as follows:
edges between the same rows:
R(x，y)＝|C NA (x 1 ，y 1 )-C NB (x 1 ，y 1 )|+|C NA (x 2 ，y 1 )-C NB (x 2 ，y 1 )| (1)
edges between different rows:
S(x，y)＝|C NA (x 1 ，y 1 )-C NB (x 1 ，y 1 )|+|C NA (x 1 ，y 2 )-C NB (x 1 ，y 2 )| (2)
the outermost edges of the frame:
wherein, the value of the node is the spectrum value of the pixel point corresponding to the original image: c (C) NA (x 1, y 1) correspondence mapImage NA is in position (x 1 ，y 1 ) Spectral value at C NA (x 2 ，y 1 ) The corresponding image NA is located (x 2 ，y 1 ) Spectral value at C NA (x 1 ，y 2 ) The corresponding image NA is located (x 1 ，y 2 ) Spectral value at C NA (x p ，y q ) For image NA at position (x p ，y q ) Spectral values at; c (C) NB (x 1 ，y 1 ) The corresponding image NB is located (x 1 ，y 1 ) Spectral value at C NB (x 2 ，y 1 ) The corresponding image NB is located (x 2 ，y 1 ) Spectral value at C NB (x 1 ，y 2 ) The corresponding image NB is located (x 1 ，y 2 ) Spectral value at C NB (x p ，y q ) For image NB in position (x p ，y q ) Spectral values at.
1.5. And (3) searching the minimum cut of the undirected graph obtained in the step (1.4) by using a graph cut algorithm to obtain a rough split joint line after resolution reduction. The method comprises the following specific steps:
all edges in the undirected graph are given a non-negative weight W e Called the cost. The graph cut algorithm may break edges in the undirected graph to form two sets of edges that are not connected to each other, such a set of edges being referred to as a "cut". If a "cut" is one, the sum of all costs for its edges is minimal, then this is referred to as a "minimal cut". In this embodiment, the graph cut algorithm is specifically a graph cut. The maximum flow/minimum cut algorithm is applied in the graph cut algorithm to obtain the minimum cut in the undirected graph, and the minimum cut divides the image into two parts, so that the dividing line is the energy optimal dividing line for reducing the images NA and NB, namely the rough splicing line.
Step two: generating a buffer area based on the rough stitching line obtained in the step one, wherein the buffer area forms a banded overlapping area in the original two-scene image, and acquiring a global optimal stitching line in the banded overlapping area.
The second step is realized by the following substeps:
2.1. generating a buffer area based on the rough stitching line obtained in the step two. The method comprises the following specific steps:
after a rough stitching line of a reduced image is obtained, selecting a buffer area radius r as 17, taking buffer areas towards two sides along the normal direction of the stitching line, obtaining a strip-shaped buffer area in the reduced image NA, marking the pixel value in the buffer area range as 1, and carrying out up-sampling to expand to the original image size to obtain an image P.
As shown in fig. 3, a schematic diagram of a buffer created in a reduced image according to the obtained rough stitching line is shown. In the figure, the buffer area is represented by a middle gray stripe, the area at the upper left of the buffer area is the area where the image NA is to be selected as the pixel value, and the area at the lower right of the buffer area is the area where the image NB is to be selected as the pixel value.
2.2. And (3) obtaining the banded overlapping area of the two images formed in the original oversized file remote sensing image by the buffer area generated in the step (2.1). The method comprises the following specific steps:
the positions of the buffers in the image P are mapped onto the original image A, B one by one, the band-shaped buffer area is obtained on the image A, B, and considering that the pixel points of the positions (the end of the stitching line) of the partial buffers of the image P after up-sampling exceed the overlapping area of the original image, in this embodiment, after the image P is mapped onto the original image, the corresponding searching and searching points at the edge, which do not belong to the overlapping area of the original image, are deleted, so as to obtain the band-shaped overlapping area of the two-scene image. And searching for an optimal splice line in the range of the banded overlapping region.
2.3. Establishing an undirected graph for pixel points in the banded overlapping region;
establishing an undirected graph G based on pixel points in the banded overlapping region obtained in the step 2.2 2 (V, E), wherein the assignment rule of each node and edge is the same as the assignment rule of each node and edge in step 2.4.
2.4. Searching the undirected graph G obtained in the step 2.3 by using a graph cut algorithm 2 And (5) the minimum cut of (V, E), specifically, referring to step 2.5, obtaining the optimal splicing line of the banded overlapping region, namely the global optimal splicing line.
The corresponding locations of the rough stitching lines within the buffer are shown in FIG. 4 (a); fig. 4 (b) is a diagram of finding the corresponding position of the resulting globally optimal patch in the band-shaped overlap region within the buffer. The rough splice lines found for the first time can be seen to be rough compared to the globally optimal splice lines found for the second time. In this embodiment, the first-time rough stitching line is segmented on the original image by 10×10 pixels, and the second-time global optimal stitching line is segmented on the original image by each pixel.
And (3) based on the global optimal stitching line obtained in the step two, uniformly coloring and stitching the original two big data remote sensing images to obtain a seamless stitched big data remote sensing image.
Fig. 5 (a) shows a gray panoramic image obtained by splicing and homogenizing two big data remote sensing images; fig. 5 (b) is a schematic diagram after the selected splicing result is partially enlarged. As can be seen from the figure, there is no significant color difference at the splice of the two images, and the splice does not substantially pass through the significant feature.
In summary, from the result of two-time stitching line searching, the big data remote sensing image can search the stitching line through the optimal stitching line acquiring method of the ultra-large file remote sensing image. Unlike the traditional single graph cut algorithm for finding the splice lines, the graph cut algorithm is nested and applied, so that the number of nodes of the graph in the graph cut algorithm is greatly reduced. It should be noted that the invention is applicable to the finding of the splice lines of various sensors, and has universal applicability to different sensors.
It will be appreciated by persons skilled in the art that the foregoing description is a preferred embodiment of the invention, and is not intended to limit the invention, but rather to limit the invention to the specific embodiments described, and that modifications may be made to the technical solutions described in the foregoing embodiments, or equivalents may be substituted for elements thereof, for the purposes of those skilled in the art. Modifications, equivalents, and alternatives falling within the spirit and principles of the invention are intended to be included within the scope of the invention.
Claims (6)
1. The optimal spelling line acquisition method of the ultra-large file remote sensing image is characterized by comprising the following steps of:
step one: acquiring two ultra-large file remote sensing images with geographic coordinate references, and performing equal resolution downsampling on the overlapped area of the two ultra-large file remote sensing images to acquire a rough splicing line;
the first step is realized by the following substeps:
(1.1) calculating the geographic overlapping area of the remote sensing images of the two-scene oversized file;
(1.2) calculating the quantity of pixels in the geographic overlapping area of the remote sensing images of the two-scene oversized file and the resampling resolution of image stitching;
(1.3) performing equal resolution downsampling treatment on the overlapping area of the remote sensing images of the two-scene oversized file;
(1.4) establishing an undirected graph for pixel points of an overlapping area of the remote sensing images of the two-scene oversized file after the downsampling treatment;
(1.5) searching the minimum cut of the undirected graph obtained in the step (1.4) by using a graph cut algorithm to obtain a rough split joint line after resolution reduction;
step two: generating a buffer area based on the rough spelling line obtained in the first step, wherein the buffer area forms a banded overlapping area in the remote sensing image of the original two-scene oversized file, and the global optimal spelling line is obtained in the banded overlapping area;
the second step is specifically realized by the following substeps:
(2.1) generating a buffer based on the rough stitching line obtained in step one;
(2.2) obtaining a banded overlapping region of the two-scene oversized file remote sensing image formed in the original oversized file remote sensing image by the buffer area generated in the step (2.1);
(2.3) establishing an undirected graph for pixel points in the banded overlapping region;
and (2.4) searching the minimum cut of the undirected graph obtained in the step (2.3) by using a graph cut algorithm, and obtaining the optimal splicing line of the banded overlapping region, namely the global optimal splicing line.
2. The method for obtaining an optimal stitching line of a remote sensing image of an oversized document according to claim 1, wherein the step (1.3) includes:
sampling coefficients n, n >1 are selected, and the two-scene oversized file remote sensing image A, B in the first step is subjected to downsampling by a bilinear interpolation method to obtain two reduced images NA and NB which are reduced by the same proportion compared with A, B, and the two reduced images NA and NB are used as images for searching rough spelling lines.
3. The method for obtaining an optimal stitching line of a remote sensing image of an oversized document according to claim 2, wherein the step (1.4) includes:
determining the range of the overlapping area on the contracted images NA and NB, and establishing an undirected graph G with weight information on each side based on the pixel points in the overlapping area 1 (V, E) the undirected graph G 1 (V, E) includes node V { C (x) 1 ，y 1 )，C(x 1 ，y 2 )，…，C(x 1 ，y t )，…，C(x k ，y t ) -and edges E { R (L), S (L), B (L) }; wherein C (x) 1 ，y 1 ) Is expressed in the position (x 1 ，y 1 ) Spectral values at C (x 1 ，y 2 ) Represented at (x 1 ，y 2 ) Spectral values at C (x 1 ，y t ) Represented at (x 1 ，y t ) Spectral values at C (x k ，y t ) Represented at (x k ，y t ) Spectral values at; r (L) represents edges between the same rows of internal nodes, S (L) represents edges between different rows of internal nodes, and B (L) represents the outermost edge of the frame;
undirected graph G 1 The assignment rules for each node and edge in (V, E) are as follows:
edges between the same rows:
R(x，y)＝|C NA (x 1 ，y 1 )-C NB (x 1 ，y 1 )|+|C NA (x 2 ，y 1 )-C NB (x 2 ，y 1 )| (1)
edges between different rows:
S(x，y)＝|C NA (x 1 ，y 1 )-C NB (x 1 ，y 1 )|+|C NA (x 1 ，y 2 )-C NB (x 1 ，y 2 )| (2)
the outermost edges of the frame:
in the formula, the value of the node is the spectrum value of the pixel point corresponding to the remote sensing image of the original oversized file: c (C) NA (x 1 ，y 1 ) For image NA at position (x 1 ，y 1 ) Spectral value at C NA (x 2 ，y 1 ) For image NA at position (x 2 ，y 1 ) Spectral value at C NA (x 1 ，y 2 ) For image NA at position (x 1 ，y 2 ) Spectral value at C NA (x p ，y q ) For image NA at position (x p ，y q ) Spectral values at; c (C) NB (x 1 ，y 1 ) For image NB in position (x 1 ，y 1 ) Spectral value at C NB (x 2 ，y 1 ) For image NB in position (x 2 ，y 1 ) Spectral value at C NB (x 1 ，y 2 ) For image NB in position (x 1 ，y 2 ) Spectral value at C NB (x p ，y q ) For image NB in position (x p ，y q ) Spectral values at.
4. The method for obtaining an optimal stitching line for a remote sensing image of an oversized document according to claim 3, wherein the step (2.1) comprises:
after the rough splicing line is obtained, selecting a buffer area with the radius r, taking buffer areas towards two sides along the normal direction of the rough splicing line, and up-sampling and expanding the buffer areas to the original image size to obtain an image P.
5. The method for obtaining an optimal stitching line for a remote sensing image of an oversized document according to claim 4, wherein the step (2.2) includes:
the positions of the buffers in the image P are mapped onto the original image A, B one by one, a banded overlapping region is obtained on the image A, B, and an optimal stitching line is found in the banded overlapping region.
6. The method for obtaining an optimal stitching line for a remote sensing image of an oversized document according to claim 5, wherein the step (2.3) includes:
establishing an undirected graph G based on pixel points in the banded overlapping region 2 (V, E) wherein the assignment rules for each node and edge are the same as the assignment rules for each node and edge in step (2.4).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211104116.0A CN116309034A (en) | 2022-09-09 | 2022-09-09 | Optimal spelling line acquisition method for ultra-large file remote sensing image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211104116.0A CN116309034A (en) | 2022-09-09 | 2022-09-09 | Optimal spelling line acquisition method for ultra-large file remote sensing image |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116309034A true CN116309034A (en) | 2023-06-23 |
Family
ID=86836428
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211104116.0A Pending CN116309034A (en) | 2022-09-09 | 2022-09-09 | Optimal spelling line acquisition method for ultra-large file remote sensing image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116309034A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116503742A (en) * | 2023-06-26 | 2023-07-28 | 自然资源部第二海洋研究所 | Remote sensing image cloud replacement method based on super-pixel and graph cut algorithm |
-
2022
- 2022-09-09 CN CN202211104116.0A patent/CN116309034A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116503742A (en) * | 2023-06-26 | 2023-07-28 | 自然资源部第二海洋研究所 | Remote sensing image cloud replacement method based on super-pixel and graph cut algorithm |
| CN116503742B (en) * | 2023-06-26 | 2023-09-08 | 自然资源部第二海洋研究所 | Remote sensing image cloud replacement method based on super-pixel and graph cut algorithm |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR101820349B1 (en) | Picture presentation method and apparatus | |
| CN108446694B (en) | Target detection method and device | |
| CN107330885B (en) | Multi-operator image redirection method for keeping aspect ratio of important content area | |
| JPH1023452A (en) | Picture extracting device and its method | |
| CN103632359A (en) | Super-resolution processing method for videos | |
| CN115546027B (en) | Image suture line determination method, device and storage medium | |
| CN106846279B (en) | A kind of adapting to image method for repairing and mending and its system based on interpolation by continued-fractions technology | |
| CN111553841B (en) | Real-time video splicing method based on optimal suture line updating | |
| CN113744142B (en) | Image restoration method, electronic device and storage medium | |
| CN112598735B (en) | Single image object pose estimation method integrating three-dimensional model information | |
| CN111105452A (en) | High-low resolution fusion stereo matching method based on binocular vision | |
| CN112288788A (en) | Monocular image depth estimation method | |
| CN111696059B (en) | Lane line smooth connection processing method and device | |
| CN116309034A (en) | Optimal spelling line acquisition method for ultra-large file remote sensing image | |
| CN111553845A (en) | Rapid image splicing method based on optimized three-dimensional reconstruction | |
| CN114998412B (en) | Shadow region parallax calculation method and system based on depth network and binocular vision | |
| CN116385288A (en) | Depth image restoration method and device and readable storage medium | |
| CN112258561A (en) | Matching point acquisition method for image stitching | |
| CN111630569B (en) | Binocular matching method, visual imaging device and device with storage function | |
| CN104392436B (en) | A kind of remote sensing image processing method and device | |
| CN111127474B (en) | Airborne LiDAR point cloud assisted orthophoto mosaic line automatic selection method and system | |
| JP4977448B2 (en) | Method and system for generating polygon boundary definitions for image objects | |
| CN110188756B (en) | Product positioning method | |
| KR100513790B1 (en) | Mapping Method of Satellite Overlay Image Map using Landsat-7 Image with SLC-off | |
| CN113034601A (en) | Scene map point and image frame matching method in environment modeling |
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 |