CN101021867A - Image searching method based on image pyramid intermediate layer blocking - Google Patents
Image searching method based on image pyramid intermediate layer blocking Download PDFInfo
- Publication number
- CN101021867A CN101021867A CN 200710048691 CN200710048691A CN101021867A CN 101021867 A CN101021867 A CN 101021867A CN 200710048691 CN200710048691 CN 200710048691 CN 200710048691 A CN200710048691 A CN 200710048691A CN 101021867 A CN101021867 A CN 101021867A
- Authority
- CN
- China
- Prior art keywords
- image
- middle layer
- image pyramid
- euclidean distance
- method based
- 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
Images
Abstract
This invention discloses an image indexing method for sub-blocks based on image pyramid interface layer, which first of all sets up an image pyramid interface layer in an original image, then unblocks the layer and then computes the brightness histogram of each sub-block and the Euclidean distance between two images and finally outputs similar image sets based on the total Euclidean distance sequence.
Description
Technical field
The present invention relates to image search method, be specifically related to a kind of image search method based on image pyramid middle layer piecemeal.
Background technology
Traditional image indexing system is when being described picture material, directly from view data, analyze mostly and extract the bottom visual signature, the color of image for example, shape, texture and spatial relation or the like, its difficult problem that faces is the relation that is difficult to accurately describe between the image content features, even accurately described the relation between the image content features, because the data volume of digital picture is big, work repetition rate height, calculating strength is big, in information retrieval based on contents, to produce more intermediate data, expend the more time, and reduced the accuracy rate of retrieval, recall ratio and rapidity.
Summary of the invention
Technical matters to be solved by this invention is how a kind of image search method based on image pyramid middle layer piecemeal is provided, this method is applicable to the image retrieval of all size size, with respect to classic method is to have utilized a plurality of content characteristics of image in retrieving simultaneously, as color, shape and spatial relation etc., its calculated amount is little, precision ratio and recall ratio height, real-time is good.
Technical matters proposed by the invention is to solve like this: a kind of image search method based on image pyramid middle layer piecemeal is provided, it is characterized in that, may further comprise the steps:
A, utilize the method for smoothly dwindling to make up the image pyramid middle layer respectively to retrieving images and the image that is retrieved;
B, piecemeal is carried out in the image pyramid middle layer;
C, calculate the brightness histogram of each sub-piece respectively;
D, calculate retrieving images and the Euclidean distance between the image of being retrieved;
Total E, according to Euclidean distance ordering, the set of output similar image, Euclidean distance is by from small to large ordering, be exactly come the most similar image of top image.
According to the image search method based on image pyramid middle layer piecemeal provided by the present invention, it is characterized in that, described image pyramid middle layer is to be the image pyramid bottom with the original image, and original image constitutes the last layer image through smoothly dwindling, by that analogy, the n tomographic image is smoothly to dwindle the image of formation by (n-1) tomographic image, by 1,2,3, ..., (n-1) layer has just made up the image pyramid middle layer.
According to the image search method based on image pyramid middle layer piecemeal provided by the present invention, it is characterized in that, the described method of smoothly dwindling makes up the image pyramid middle layer and can adopt weights to equate that method makes up the image pyramid middle layer and comprises step: get the image that contains plurality of pixels point, be used as the pixel value of a new pixel with the weighted mean value of this plurality of pixels value, by that analogy, just constitute the level and smooth downscaled images of upper level image by all new pixels of next stage, obtained the middle layer that image through smoothly dwindling has just constituted image pyramid step by step.
According to the image search method based on image pyramid middle layer piecemeal provided by the present invention, it is characterized in that, in described step B, get pixel value in the image pyramid middle layer and be 256 * 256 figure layer, be divided into 16 * 16 totally 256 sub-pieces, every is 16 * 16 pixels.
According to the image search method based on image pyramid middle layer piecemeal provided by the present invention, it is characterized in that: at first calculate retrieving images and the corresponding sub-Block Brightness histogram of the image that is retrieved between Euclidean distance, again the Euclidean distance addition of all sub-pieces is obtained total Euclidean distance.
Beneficial effect of the present invention is: this method is that calculated amount is little with respect to other method, precision ratio and recall ratio height, and real-time is good.Owing to used the image pyramid middle layer, piecemeal and corresponding sub-piece matching strategy carried out in the middle layer, be applicable to the image of all size size.Basic thought is to dwindle original image, reduces calculated amount.Fully utilized the content characteristics such as color, shape and spatial relation of original image simultaneously, therefore had higher precision ratio and recall ratio, real-time is good.
Description of drawings
Fig. 1 is an image pyramid interlayer structure synoptic diagram provided by the present invention;
Fig. 2 is the synoptic diagram of an embodiment of the method for smoothly dwindling;
Fig. 3 is the synoptic diagram to an embodiment of image middle layer piecemeal.
Embodiment
The present invention is described further below in conjunction with accompanying drawing and embodiment.
Based on the image search method of image pyramid middle layer piecemeal, at first, from original image, make up the image pyramid middle layer.As shown in Figure 11,2,3 ..., (n-1) layer is defined as the image pyramid middle layer, and 0, the n layer is defined as image pyramid bottom and top layer respectively.Wherein 0 layer is original image, and 1 layer is smoothly to dwindle the image that constitutes by 0 layer of process, by that analogy, the n layer be by (n-1) layer through smoothly dwindling the image of formation, by 1,2,3 ..., (n-1) layer has just made up the image pyramid middle layer.Smoothly dwindling method, is to get the image that contains m pixel, is used as the pixel value of a new pixel with the weighted mean value of m some pixel value, and method like this just has been made of the level and smooth downscaled images of upper level image all new pixels of next stage.Obtain the middle layer that image through smoothly dwindling has just constituted image pyramid step by step.A kind of concrete grammar that smoothly dwindles as shown in Figure 2,1 to 9 occupied 9 pixels have only 1 pixel among the figure after smoothly dwindling, the position is at 5 places, value is the weighted mean value of these 9 points, has weights not wait smoothly dwindling of method in addition, repeats no more.Get certain one-level in image pyramid middle layer during image retrieval, its pixel value lacks than original image, generally gets about 256 * 256, can be by the suitable convergent-divergent of original image size.From construction method as can be known, all images pyramid middle layer all derives from original image, is the epitome of the different sizes of original image, has not only comprised the content characteristics such as color, texture, shape and spatial relation of original image, and pixel is few, data volume is little, is convenient to calculate.
Secondly, piecemeal is carried out in the image pyramid middle layer.Desirable about 256 * 256 sizes in general pattern pyramid middle layer are divided into 16 * 16 totally 256 sub-pieces, and every is 16 * 16 pixels.A kind of concrete grammar of image block as shown in Figure 3, it is the piecemeal synoptic diagram, just can be divided into 256, ask the Euclidean distance between the corresponding blocks respectively, the Euclidean distance of suing for peace always more at last, distance heals little more similar, this also is one of partition strategy, also can be divided into 256 * 256, so long as 2 exponential just passable, mathematical operation is convenient like this.
Then, calculate the brightness histogram of each sub-piece respectively.For the pixel of each sub-piece, R, G, three pixel values of B are arranged, brightness-formula is shown in (1).Can obtain the brightness value Y of this point by (1).Because R, G, three pixel values of B all are not more than 255, have so just determined one 256 grades gray scale,, calculate brightness histogram and reduced by 2/3rds calculated amount with respect to the color histogram of R, G, three pixel values of B.
Y＝0.299R+0.587G+0.114B (1)
Then, calculate Euclidean distance between two width of cloth images.This two width of cloth image refers to retrieving images U and the image V that is retrieved, and they obtain brightness histogram by Same Way.At first calculate the Euclidean distance between the corresponding sub-Block Brightness histogram of two width of cloth images, again the Euclidean distance addition of all sub-pieces is obtained total Euclidean distance.Euclidean distance D between the corresponding sub-piece i
iComputing method as shown in Equation (2):
U wherein
j, V
jIt is respectively the characteristic quantity that extracts from retrieving images U and the corresponding sub-Block Brightness histogram of the image V that is retrieved.The computing method of total Euclidean distance D are as shown in Equation (3):
Wherein n is the maximum piecemeal subnumber in the image pyramid middle layer of being got, as presses piecemeal shown in the table 2, and its value is 256.
At last, according to total Euclidean distance ordering, the set of output similar image.When the Euclidean distance minimum was 0, this two width of cloth image had maximum similarity 100%.Euclidean distance just is equivalent to similarity by arranging from big to small by arranging from small to large, and coming top image is exactly the most similar image.
Claims (5)
1, a kind of image search method based on image pyramid middle layer piecemeal is characterized in that, may further comprise the steps:
A, utilize the method for smoothly dwindling to make up the image pyramid middle layer respectively to retrieving images and the image that is retrieved;
B, piecemeal is carried out in the image pyramid middle layer;
C, calculate the brightness histogram of each sub-piece respectively;
D, calculate retrieving images and the Euclidean distance between the image of being retrieved;
Total E, according to Euclidean distance ordering, the set of output similar image, Euclidean distance is by ordering from small to large, be exactly come the most similar image of top image.
2, the image search method based on image pyramid middle layer piecemeal according to claim 1, it is characterized in that, described image pyramid middle layer is to be the image pyramid bottom with the original image, and original image constitutes the last layer image through smoothly dwindling, by that analogy, the n tomographic image is smoothly to dwindle the image of formation by (n-1) tomographic image, by 1,2,3, ..., (n-1) layer has just made up the image pyramid middle layer.
3, image search method based on image pyramid middle layer piecemeal according to claim 1, it is characterized in that, the described method of smoothly dwindling makes up the image pyramid middle layer and can adopt weights to equate that method makes up the image pyramid middle layer and comprises step: get the image that contains plurality of pixels point, be used as the pixel value of a new pixel with the weighted mean value of this plurality of pixels value, by that analogy, just constitute the level and smooth downscaled images of upper level image by all new pixels of next stage, obtained the middle layer that image through smoothly dwindling has just constituted image pyramid step by step.
4, the image search method based on image pyramid middle layer piecemeal according to claim 1 is characterized in that, in described step B, get the image pyramid middle layer, as pixel value is 256 * 256 middle graph layer, can be divided into 16 * 16 totally 256 sub-pieces, and every is 16 * 16 pixels.
5, the image search method based on image pyramid middle layer piecemeal according to claim 1, it is characterized in that: at first calculate retrieving images and the corresponding sub-Block Brightness histogram of the image that is retrieved between Euclidean distance, again the Euclidean distance addition of all sub-pieces is obtained total Euclidean distance.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 200710048691 CN101021867A (en) | 2007-03-22 | 2007-03-22 | Image searching method based on image pyramid intermediate layer blocking |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 200710048691 CN101021867A (en) | 2007-03-22 | 2007-03-22 | Image searching method based on image pyramid intermediate layer blocking |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN101021867A true CN101021867A (en) | 2007-08-22 |
Family
ID=38709630
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 200710048691 Pending CN101021867A (en) | 2007-03-22 | 2007-03-22 | Image searching method based on image pyramid intermediate layer blocking |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101021867A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102253989A (en) * | 2011-07-04 | 2011-11-23 | 厦门市美亚柏科信息股份有限公司 | Image processing method and device, and image retrieval method and system |
| CN102403011A (en) * | 2010-09-14 | 2012-04-04 | 北京中星微电子有限公司 | Music output method and device |
| CN101739658B (en) * | 2008-11-06 | 2012-09-05 | 索尼株式会社 | Image processing apparatus, image processing method, and program |
| CN101630360B (en) * | 2008-07-14 | 2012-12-19 | 上海分维智能科技有限公司 | Method for identifying license plate in high-definition image |
| CN103477309A (en) * | 2011-02-28 | 2013-12-25 | 意法半导体(R&D)有限公司 | Improvements in or relating to optical navigation devices |
| CN108388670A (en) * | 2018-03-20 | 2018-08-10 | 华北电力大学（保定） | Quaternary tree based on local codebook divides the image search method of shape |
-
2007
- 2007-03-22 CN CN 200710048691 patent/CN101021867A/en active Pending
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101630360B (en) * | 2008-07-14 | 2012-12-19 | 上海分维智能科技有限公司 | Method for identifying license plate in high-definition image |
| CN101739658B (en) * | 2008-11-06 | 2012-09-05 | 索尼株式会社 | Image processing apparatus, image processing method, and program |
| CN102403011A (en) * | 2010-09-14 | 2012-04-04 | 北京中星微电子有限公司 | Music output method and device |
| CN103477309A (en) * | 2011-02-28 | 2013-12-25 | 意法半导体(R&D)有限公司 | Improvements in or relating to optical navigation devices |
| US9354719B2 (en) | 2011-02-28 | 2016-05-31 | Stmicroelectronics (Research & Development) Limited | Optical navigation devices |
| CN102253989A (en) * | 2011-07-04 | 2011-11-23 | 厦门市美亚柏科信息股份有限公司 | Image processing method and device, and image retrieval method and system |
| CN102253989B (en) * | 2011-07-04 | 2013-10-09 | 厦门市美亚柏科信息股份有限公司 | Image processing method and device, and image retrieval method and system |
| CN108388670A (en) * | 2018-03-20 | 2018-08-10 | 华北电力大学（保定） | Quaternary tree based on local codebook divides the image search method of shape |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110032954B (en) | Intelligent identification and counting method and system for reinforcing steel bars | |
| CN110197182A (en) | Remote sensing image semantic segmentation method based on contextual information and attention mechanism | |
| JP4486780B2 (en) | Nonlinear quantization and similarity matching method for image information retrieval | |
| CN110232394A (en) | A kind of multi-scale image semantic segmentation method | |
| CN101923707B (en) | Watershed algorithm-based high spatial resolution multi-spectral remote sensing image segmentation method | |
| CN109064396A (en) | A kind of single image super resolution ratio reconstruction method based on depth ingredient learning network | |
| CN101021867A (en) | Image searching method based on image pyramid intermediate layer blocking | |
| CN107403430A (en) | A kind of RGBD image, semantics dividing method | |
| CN105976318A (en) | Image super-resolution reconstruction method | |
| CN102567731A (en) | Extraction method for region of interest | |
| CN112862689B (en) | Image super-resolution reconstruction method and system | |
| CN110322453A (en) | 3D point cloud semantic segmentation method based on position attention and auxiliary network | |
| CN113160062B (en) | Infrared image target detection method, device, equipment and storage medium | |
| CN110148091A (en) | Neural network model and image super-resolution method based on non local attention mechanism | |
| CN110245249B (en) | Three-dimensional CAD model intelligent retrieval method based on double-layer depth residual error network | |
| CN114648684A (en) | Lightweight double-branch convolutional neural network for image target detection and detection method thereof | |
| CN111738114B (en) | Vehicle target detection method based on anchor-free accurate sampling remote sensing image | |
| CN104835196A (en) | Vehicular infrared image colorization and three-dimensional reconstruction method | |
| CN113449691A (en) | Human shape recognition system and method based on non-local attention mechanism | |
| CN107833224A (en) | A kind of image partition method based on multi-level region synthesis | |
| CN101587189A (en) | Texture elementary feature extraction method for synthetizing aperture radar images | |
| CN114898200A (en) | Conveyor belt workpiece image target detection method and system based on light-weight YOLOV4-tiny | |
| CN102737381B (en) | Image partitioning method based on mixed bipartite graph clustering integration | |
| CN103377477B (en) | A kind of high-resolution remote sensing image multilayer division method | |
| CN112270366A (en) | Micro target detection method based on self-adaptive multi-feature fusion |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C12 | Rejection of a patent application after its publication | ||
| RJ01 | Rejection of invention patent application after publication |
Open date: 20070822 |