CN105657436A - Image processing method - Google Patents
Image processing method Download PDFInfo
- Publication number
- CN105657436A CN105657436A CN201511026619.0A CN201511026619A CN105657436A CN 105657436 A CN105657436 A CN 105657436A CN 201511026619 A CN201511026619 A CN 201511026619A CN 105657436 A CN105657436 A CN 105657436A
- Authority
- CN
- China
- Prior art keywords
- image
- small object
- compression
- decoding
- global
- 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
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/60—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding
- H04N19/63—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding using sub-band based transform, e.g. wavelets
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/90—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using coding techniques not provided for in groups H04N19/10-H04N19/85, e.g. fractals
- H04N19/96—Tree coding, e.g. quad-tree coding
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Compression Of Band Width Or Redundancy In Fax (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention provides an image processing method. The image processing method comprises the following steps: performing target extraction according to a global image, and acquiring a small target image, wherein the small target image comprises a small target image position and a small target image pixel; compressing the global image and the small target image respectively to obtain a compressed global image and a compressed small target image; encoding the compressed global image and the compressed small target image to synthesize an encoding result; and receiving the encoding result, performing inverse transformation decoding on a compressed global image code and a compressed small target image code to obtain a decoded global image and a decoded small target image, performing decoded small target image pixel replacement on a same position of the decoded global image according to the decoded small target position, and performing image recovery to obtain a remote sensing image, wherein the decoded small target image comprises a decoded small target position and a decoded small target image pixel.
Description
Technical field
The present invention relates to technical field of image processing, particularly to the method for the Small object compression of images of the remote sensing images in image procossing.
Background technology
Along with the development of space technology and remote sensing technology, remote sensing images obtain increasingly be widely applied at life, production and aerospace fields. but, along with the raising of satellite borne sensor resolution, the application of multispectral, high-spectrum remote-sensing, the acquisition rate of remote sensing images and gather the throughput rate that capacity transmits considerably beyond data, Remote Sensing Image Compression becomes requisite approach. in recent years, Remote Sensing Image Compression technology achieves development at full speed, currently used it is based on wavelet transformation more widely, based on Design Based on Spatial Resampling and based on structure describe remote sensing image compression method, based on the compress technique of wavelet transformation at compression of images field gradual perfection, it has been applied in MPEG-4 and JPEG2000 compression standard, and the compression algorithm (compression ratio is about 4) when relatively low compression ratio based on Design Based on Spatial Resampling has good compression quality, compression algorithm other algorithms relative on compression quality described based on structure have and are obviously improved.
High spatial resolution remote sensing towards spaceborne application compresses, and these three method exists a maximum shortcoming: be easily lost in image the minutia of some Small object in compression process, but this information is important again, often has important applying value.
Although small echo has the approximation capability of optimum in representing the some singularity of signal; it has had bigger application in compression of images field; but due to wavelet transformation deficiency in directivity and anisotropy, cause that JPEG2000 can not protect the geological information such as image border and texture effectively. Despite the compact schemes characteristic decomposing realization and its basic function with the Multilevel Iteration of wavelet transformation so that it can take into account the rarefaction representation of energy concentration and singular point feature in compression of images is applied. But it is also noted that, in special application fields such as High spatial resolution remote sensing compressions, wavelet transformation still has the disadvantage that in the coefficient matrix that (1) multilevel wavelet decomposes, basic function is long, in low bit rate situation, the image block of compression there will be more serious diplopia phenomenon, also create the fuzzy and isolated speckle of the edge pollution near zone in image simultaneously, cause pattern distortion.(2) such as conventional wavelet scaling function changes by binary system, therefore poor in its frequency resolution of HFS, and remote sensing images medium-high frequency abundant information, and generally corresponds to important goal. (3) two-dimensional wavelet transformation realized by tensor product can not effectively represent have directive edge feature target in image. Additionally the algorithm flow of wavelet transformation also determine its compression method Small object information keep on there is certain inferior position: first, based on the information that can give up a part of frequency-domain segment in the method for compressing image of wavelet transformation, and the frequency-domain segment giving up some part is possible to cause the loss of Small object information. Based in the image compression process of wavelet transformation, in order to ensure certain image compression rate, after every one-level wavelet transformation, all only retain some frequency-domain segment it is carried out the iteration of next stage wavelet transformation. Secondly, the quality of compression of images also can be produced impact by some defects of wavelet transformation self, especially the long meeting of basic function of coefficient matrix that multilevel wavelet decomposes causes in the compression of images of low bit rate occurs that more serious diplopia phenomenon or image border place cause pattern distortion, conventional wavelet scaling function also results in it be deteriorated in the frequency resolution of HFS by what binary system changed, affects the compression quality of high-frequency information in Remote Sensing Image Compression features such as () edge of such as image, projections. Other compression algorithm is such as less practically applicable to the application of high compression ratio based on the compression algorithm of Design Based on Spatial Resampling, and when compression ratio promotes, its compression quality can be subject to large effect; Based on structure describe mode for texture classified description can't overlay image institute textured.
Therefore, how providing a kind of novel transformation theory represent and process the High dimensional space datas such as image, the New Image compression algorithm with target retention performance becomes the technical problem that the lifting of current image compression apparatus faces.
Summary of the invention
For the problems referred to above that prior art exists, present invention improves over the method improving Small object definition by Optimize Compression Algorithm of prior art, provide and a kind of carry out individually lossless or small reduction ratio compression for Small object region, purpose is in that effectively to solve previous methods distortion on Small object keeps or the deficiency of material particular information dropout, improves Remote Sensing Image Compression in the application to occasions such as Small object information fidelity are more sensitive such as small target tracking, Small object investigation.
For achieving the above object, this technological invention provides techniques below scheme:
Carrying out Objective extraction according to global image, obtain Small object image, described Small object image includes Small object picture position and Small object image pixel; Respectively global image and Small object image are compressed, obtain compression global image and compression Small object image; Compression global image and compression Small object image are encoded, composite coding result; Receive coding result, compression global image coding and compression Small object picture coding are carried out inverse transformation decoding, obtain decoding global image and decoding Small object image, described decoding Small object image includes decoding Small object position and decoding Small object image pixel, it is decoded the same position of decoding global image after Small object image pixel replaces it, carrying out image and recovering to obtain remote sensing images according to decoding Small object position.
Further, the extracting method of described Small object image, for using the notable graph model of the overall situation to carry out Small object extraction, first goes out Small object region in background image according to the significance model inspection of remote sensing hyperspectral image;Then Pixel Information and the positional information in Small object region are extracted.
Further, the process of described compression global image, it is compressed by 3 grades of traditional wavelet transformations of main employing, finally extracts the information of " LL " the layer frequency domain part after 3rd level wavelet transformation, to retain the texture of background image, global information.
Further, processing as it being carried out lossless or small reduction ratio compression according to the size adaptation in its region of described compression Small object image, the method for described compression Small object image is: when Small object region, and the compression ratio chosen time big is big; The compression ratio chosen when Small object region is little is little; It is 0 when Small object region is the compression ratio chosen without compression.
Further, the coded system that method is ADPCM (DPMC) of described compression coding global image, compression global image obtains compression global image coding after carrying out DPMC coding, compression global image coding can obtain global image information by DPMC inverse transformation.
Further, the coded system that method is Embedded Wavelet zero tree (EZW) of described compression coding Small object image, compression Small object image obtains compression Small object picture coding after carrying out EZW coding, compression Small object picture coding can obtain Small object image information by EZW inverse transformation.
Further, the image recovery method of described decoding global image is wavelet inverse transformation.
In sum, present invention employs the small target detecting method of the remote sensing images based on spectrum saliency and analysis of neural network identification, the target differed greatly with ambient background can be detected and using the significant main constituent of spectrum as the input of neutral net what there is no a priori, solve target moving, rotate and impalpable problem in process that shape is continually changing or under ambiguous state. The method effectively improves the robustness identifying Small object and accuracy, especially improves Small object Tracking Recognition effect under special state.
The present invention proposes the compression processing mode in a kind of individual processing Small object region, owing to Small object region will not be very big, carry out taking very big bandwidth when lossless or small reduction ratio compression processes and do not result in data transmission to it, but being different from the conventional way being improved Small object details fidelity by Optimize Compression Algorithm, the present invention is greatly improved the definition of the Small object minutia that image recovers.
The present invention proposes a kind of self adaptation and chooses the method that Small object region is processed by compression ratio, quantity and the area size of the Small object processed as required choose compression ratio, lossless compress is adopted when quantity is few, region is little, adopt small reduction ratio to be compressed when quantity is more, region is bigger, suitable reduces bandwidth required during image transmitting, the adaptability of compression algorithm can be greatly improved, the high true contradiction between property and image transmitting bandwidth of Optimized Matching Small object.
Accompanying drawing explanation
Accompanying drawing is used to provide a further understanding of the present invention, and constitutes the part of description, is used for explaining the present invention, but is not intended that limitation of the present invention together with detailed description below. In the accompanying drawings:
Fig. 1 is the flow chart of image processing method in the embodiment of the present invention
Fig. 2 is the schematic diagram that in the embodiment of the present invention, image Small object extracts
Fig. 3 is the flow chart of global image compression in the embodiment of the present invention
Fig. 4 is the schematic diagram of the extraction of image, coding in the embodiment of the present invention
Fig. 5 is the structure buffer memory schematic diagram encoding result in the embodiment of the present invention
Fig. 6 is the schematic diagram encoding the picture decoding of result, recovery in the embodiment of the present invention
Detailed description of the invention
Below in conjunction with accompanying drawing, the specific embodiment of the present invention is described in detail.It should be appreciated that detailed description of the invention described herein is merely to illustrate and explains the present invention, it is not limited to the present invention. Therefore, those of ordinary skill in the art will be appreciated that, it is possible to the embodiment of description here is made various change and amendment, without departing from scope and spirit of the present invention. Equally, for clarity and conciseness, description below eliminates the description to known function and structure.
In order to improve the definition of image procossing minutia, embodiments of the invention provide a kind of method of image procossing.
As it is shown in figure 1, image processing method provided by the invention, comprise the following steps:
Step 101, carries out Objective extraction according to global image, obtains Small object image, and described Small object image includes Small object picture position and Small object image pixel;
Step 102, is compressed global image and Small object image respectively, obtains compression global image and compression Small object image;
Step 103, is encoded compression global image and compression Small object image, composite coding result;
Step 104, receive coding result, compression global image coding and compression Small object picture coding are carried out inverse transformation decoding, obtain decoding global image and decoding Small object image, described decoding Small object image includes decoding Small object picture position and decoding Small object image pixel, it is decoded carrying out image recovery after Small object image pixel replaces it to the same position of decoding global image according to decoding Small object picture position, obtains remote sensing images.
As in figure 2 it is shown, the method for the embodiment of the present invention includes, global image is carried out small target deteection, extract position and the pixel of Small object. Data are also processed while comprising great amount of images information and bring bigger difficulty by high-spectral data, so requiring over principal component analysis high-spectral data is carried out dimension-reduction treatment, forming component abstraction sequence after main constituent rotates through, takes the input value comprising the bigger sequential value of composition information as neutral net. By error backpropagation algorithm, the weights of neutral net are trained, make the error sum of squares between neutral net output and actual output reach expected value.
In order to improve the high-resolution of compression of images Small Target, scheme extracts Small object and carries out individual processing. The present invention adopts the overall situation spectrum notable figure Small object to detect in remote sensing images, and the formula that is expressed as of the significance of each pixel in image is: (x, y)=�� mi, wherein i value is 0 to count to the maximum pixel of remote sensing images to Sg '.
As shown in Figure 3, the global image boil down to of the embodiment of the present invention, adopt 3 grades of traditional wavelet transformations that it is compressed, finally extract the information of " LL " the layer frequency domain part after 3rd level wavelet transformation, the compression method of this embodiment can retain its general picture, textural characteristics as much as possible, and therefore it is carried out next stage iteration by the LL section frequency domain of the every one-level wavelet transformation of scheme reservation.
As shown in Figure 4, the combined coding that picture coding mode is Small object region and global image of the embodiment of the present invention. After processing complete primary school target area and global image, it is simply that it is encoded transmission. Owing to taking processing method two kinds different for Small object with global image, both data take on a different character. Data after Small object process are mainly made up of pixel value, and its data have average feature big, that change is little relatively, and the data of global image are the values after 3 grades of wavelet transformations, and its data change greatly relatively.Therefore combined coding as shown in Figure 4 is for the view data in Small object region, take ADPCM (DPCM) coded system that method is simple and is easily achieved, for global image data, take the coded system of " Embedded Wavelet coefficient zero tree Image Coding (EZW) ".
As it is shown in figure 5, the picture coding of the embodiment of the present invention synthesizes after encoding result, carry out structure buffer memory and can carry out data transmission. Picture coding result is arranged in order the value after into the positional information of Small object, Small object pixel value and global image wavelet transformation, in order to distinguish each several part data, have employed a kind of format header and indicate the data of different section, Fig. 5 indicates I Hearder, II Hearder, III Hearder represent the positional information of Small object respectively, the original position of the pixel value of Small object and global image wavelet transform result. The positional information of Small object is expressed as that (x, y, l, w), wherein (x, y) represents the coordinate position in the upper left corner, Small object region, and (l w) represents the annotation length of Small object rectangle frame and width.
As shown in Figure 6, the coding result of the embodiment of the present invention is replaced by Small object and is carried out image recovery with small echo inversion. The coding result treatment of this embodiment is divided into following 3 steps: first extract the data of the positional information of Small object, Small object pixel value and three parts of global image wavelet transform result from the data buffer storage structure shown in Fig. 5; Then according to Small object information and wavelet transform result are decoded by the inverse transformation of DPCM and EZW respectively; In the global image that EZW decoding obtains, finally merge the information of Small object exactly. Owing to can be obtained the positional information (x in Small object region by decoding, y, l, w), therefore the Small object pixel value obtained with decoding replaces the pixel in global image Small Target region, it is possible to increase exponentially the definition of the remote sensing images Small Target of recovery.
The image processing method of the present invention is in the application of this embodiment, and image can carry out buffer memory or data transmission through the coding result extracted, compress and obtain after coding. Being mainly used in the data transmission between remote sensing equipment and local device, remote sensing equipment is used for catching picture signal, for instance: video, obtain imaging clearly by image procossing at local device end.
Obviously, the present invention can be carried out various change and modification without deviating from the spirit and scope of the present invention by those skilled in the art. So, if these amendments of the present invention and modification belong within the scope of the claims in the present invention and equivalent technologies thereof, then the present invention is also intended to comprise these change and modification.
Claims (9)
1. an image processing method, it is characterised in that comprise the following steps:
(1) carrying out Objective extraction according to global image, obtain Small object image, described Small object image includes Small object picture position and Small object image pixel;
(2) respectively global image and Small object image are compressed, obtain compression global image and compression Small object image;
(3) compression global image and compression Small object image are encoded, composite coding result;
(4) coding result is received, compression global image coding and compression Small object picture coding are carried out inverse transformation decoding, obtain decoding global image and decoding Small object image, described decoding Small object image includes decoding Small object picture position and decoding Small object image pixel, it is decoded carrying out image recovery after Small object image pixel replaces it to the same position of decoding global image according to decoding Small object picture position, obtains remote sensing images.
2. a kind of image processing method according to claim 1, it is characterised in that the extracting method of described Small object image carries out Small object extraction for using the notable graph model of the overall situation.
3. a kind of image processing method according to claim 1, it is characterised in that the method for described compression global image is: three grades of wavelet transformations.
4. a kind of image processing method according to claim 1, it is characterised in that the method for described compression Small object image is: when Small object region, the compression ratio chosen time big is big; The compression ratio chosen when Small object region is little is little; It is 0 when Small object region is the compression ratio chosen without compression.
5. a kind of image processing method according to claim 1, it is characterised in that the method for described compression coding global image is: ADPCM.
6. a kind of image processing method according to claim 1, it is characterised in that the method for described compression coding Small object image is: described compression Small object method for encoding images is Embedded Wavelet zero tree.
7. a kind of image processing method according to claim 1, it is characterised in that the method for described inverse transformation decoding compression global image coding is: ADPCM inverse transformation.
8. a kind of image processing method according to claim 1, it is characterised in that the method for described inverse transformation decoding compression Small object picture coding is: Embedded Wavelet zero tree inverse transformation.
9. a kind of image processing method according to claim 1, it is characterised in that the image recovery method of described decoding global image is: wavelet inverse transformation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201511026619.0A CN105657436A (en) | 2015-12-31 | 2015-12-31 | Image processing method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201511026619.0A CN105657436A (en) | 2015-12-31 | 2015-12-31 | Image processing method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105657436A true CN105657436A (en) | 2016-06-08 |
Family
ID=56491084
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201511026619.0A Pending CN105657436A (en) | 2015-12-31 | 2015-12-31 | Image processing method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105657436A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111369457A (en) * | 2020-02-28 | 2020-07-03 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Remote sensing image denoising method for sparse discrimination tensor robustness PCA |
| CN113039576A (en) * | 2018-11-15 | 2021-06-25 | 苹果公司 | Image enhancement system and method |
-
2015
- 2015-12-31 CN CN201511026619.0A patent/CN105657436A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113039576A (en) * | 2018-11-15 | 2021-06-25 | 苹果公司 | Image enhancement system and method |
| CN111369457A (en) * | 2020-02-28 | 2020-07-03 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Remote sensing image denoising method for sparse discrimination tensor robustness PCA |
| CN111369457B (en) * | 2020-02-28 | 2022-05-17 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Remote sensing image denoising method for sparse discrimination tensor robustness PCA |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110225341A (en) | A kind of code flow structure image encoding method of task-driven | |
| CN102186069B (en) | Remote sensing image data compression method capable of maintaining measurement performance | |
| CN104199627B (en) | Gradable video encoding system based on multiple dimensioned online dictionary learning | |
| CN110211016B (en) | Watermark embedding method based on convolution characteristic | |
| Kuanar et al. | Gated fusion network for sao filter and inter frame prediction in versatile video coding | |
| CN105657436A (en) | Image processing method | |
| CN103544717B (en) | A kind of two-stage three dimensional image processing coded method based on SIFT feature | |
| Zhao et al. | Intelligent analysis oriented surveillance video coding | |
| Bajpai et al. | 3D modified wavelet block tree coding for hyperspectral images | |
| WO2023203509A1 (en) | Image data compression method and device using segmentation and classification | |
| CN114463449A (en) | Hyperspectral image compression method based on edge guide | |
| Avola et al. | Forward-looking sonar image compression by integrating keypoint clustering and morphological skeleton | |
| Hu et al. | Revisit Visual Representation in Analytics Taxonomy: A Compression Perspective | |
| Deng et al. | MASIC: Deep Mask Stereo Image Compression | |
| Rana et al. | Data compression algorithm for computer vision applications: A survey | |
| Bao et al. | Image Compression for Wireless Sensor Network: A Model Segmentation-Based Compressive Autoencoder | |
| CN105392015B (en) | A kind of cartoon image compression method based on explicit mixing reconciliation diffusion | |
| Hou et al. | Interband prediction compressed sensing reconstruction algorithm for hyperspectral image | |
| Kavitha et al. | Lossy compression through segmentation on low depth-of-field images | |
| Vijay et al. | A review of image transmission using real time technique over wmsn | |
| Cha et al. | Joint SAR image compression and coherent change detection | |
| Lukin et al. | Automation of Processing Multichannel Remote Sensing Images Based on Performance Prediction | |
| Zhu et al. | Hybrid scheme for efficient shape coding | |
| Wang et al. | Automatic Region-of-Interest Coding in JPEG2000 Based on Morphology Segmentation and LL n Subband Analysis | |
| Avola et al. | A forward-looking sonar progressive coding using morphological skeleton representation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20160608 |
|
| WD01 | Invention patent application deemed withdrawn after publication |