CN104835114A - Image self-adaptive display method - Google Patents
Image self-adaptive display method Download PDFInfo
- Publication number
- CN104835114A CN104835114A CN201510227064.XA CN201510227064A CN104835114A CN 104835114 A CN104835114 A CN 104835114A CN 201510227064 A CN201510227064 A CN 201510227064A CN 104835114 A CN104835114 A CN 104835114A
- Authority
- CN
- China
- Prior art keywords
- image
- remarkable
- pixel
- finedraw
- overall situation
- 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 45
- 230000000699 topical effect Effects 0.000 claims description 29
- 230000003044 adaptive effect Effects 0.000 claims description 27
- 238000005520 cutting process Methods 0.000 claims description 11
- 230000008569 process Effects 0.000 claims description 9
- 238000011156 evaluation Methods 0.000 claims description 8
- 238000012856 packing Methods 0.000 claims description 8
- 230000000007 visual effect Effects 0.000 claims description 8
- 230000008859 change Effects 0.000 claims description 7
- 239000011159 matrix material Substances 0.000 claims description 4
- 238000005259 measurement Methods 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 230000008447 perception Effects 0.000 claims description 3
- 238000012544 monitoring process Methods 0.000 abstract 1
- 230000000149 penetrating effect Effects 0.000 abstract 1
- 238000001303 quality assessment method Methods 0.000 abstract 1
- 238000013459 approach Methods 0.000 description 9
- 238000010586 diagram Methods 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 4
- 238000013507 mapping Methods 0.000 description 4
- 238000013441 quality evaluation Methods 0.000 description 4
- 230000008878 coupling Effects 0.000 description 2
- 238000010168 coupling process Methods 0.000 description 2
- 238000005859 coupling reaction Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 235000002566 Capsicum Nutrition 0.000 description 1
- 240000008574 Capsicum frutescens Species 0.000 description 1
- 239000001390 capsicum minimum Substances 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000010008 shearing Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000017105 transposition Effects 0.000 description 1
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
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
Abstract
The invention relates to an image self-adaptive display method combining slit cut and uniform scaling based on image content awareness. Firstly, a saliency map integrating local and global significance is used for replacing a gradient map in traditional slit cut to prevent a slit to be cut from penetrating a low-energy region of interest (ROI) so that the geometry distortion of an important object can be avoided. Then, a slit cut method and a uniform scaling method are combined to minimize the difference between an original image and an image after self-adaptation. Specifically, while the slit cut method is performed, a quality assessment method is used for calculating the distance generated between a temporary image and the original image, that is real-time monitoring of deformation of important information; and when the deformation exceeds a set index, the slit cut method is terminated and switched to the uniform scaling method. Compared with the traditional slit cut, uniform scaling and other self-adaptive methods, the image self-adaptive method can not only effectively protects important regions of the image from deformation, but can also compress unimportant regions to ensure global information.
Description
Technical field
The present invention relates to a kind of for the image adaptive display packing on different size display terminal, belong to image, multimedia signal processing technique field.
Background technology
Along with popularizing of mobile device, be applicable to the continuous update of terminal device of heterogeneous networks, TV, notebook computer, PDA, mobile phone etc., existence that is dissimilar, different size display terminal becomes objective fact.In order to ensure that the user holding distinct device can both cosily watch same picture material, require that picture material can be adaptive to the user terminal of heterogeneous networks situation, different size, different proportion, this just requires that image has adaptive technique.When we are by little to screen size for high definition picture transfer large for a width size and when the display terminal that resolution is low shows, the problems such as some the Key detail region loss in anamorphose distortion or picture may be there is, thus have impact on spectators and appreciate quality.In order to make the user of distinct device can the cosily intact image not having distortion of view content, this just needs an image adaptive system accurately, enable the adaptive terminal device being shown in different size of image, make spectators have best viewing experience.
In recent years, there is increasing researcher to start to pay close attention to image adaptive and shown this problem.Traditional image adaptive method comprises even convergent-divergent (scaling) technology and simply cuts (cropping) technology.Even convergent-divergent is exactly carry out to original image the image that even down-sampling or interpolation obtain target size, and when we change the even convergent-divergent of the ratio of width to height of image, often cause the stretching distorting transformation of image important area content, and screen size is less, resolution is lower, the distortion produced is more obvious, causes so serious to view and admire Quality Down.The technology that simply cuts is the target image that the edge content manually dismissing image obtains showing the most outshot of image, and this method can cut some important information regions to different display sizes, thus cause the loss of the Key detail of some images, important area is shown imperfect, image content is incomplete.
The key issue of image adaptive display is the important content how obtaining image, and makes its important content or perpetual object region can be good at showing on the screen of different size, different resolution.A kind of conventional method is: change the size of image by unnoticed pixel content in the method process image of slit cropping; important area is protected deformation not to occur to greatest extent; retain the global information of original image as much as possible, but how to carry out evaluation to the visual impact of slit cropping be a large difficult point simultaneously.
Summary of the invention
The present invention is directed to the problems such as the loss of metamorphopsic distortion that high-definition picture exists when small size or lower resolution displays show and Key detail, provide a kind of can adaptive adjustment picture size, ensure the display of image important area information completely and the minimum image adaptive display packing for different size display terminal of image fault.The present invention adopts a kind of criteria of quality evaluation to monitor the deformation extent of important information, cutting out and evenly balancing between convergent-divergent process, stoping image important objects edge distortion by setting certain threshold value at finedraw.Present invention incorporates the remarkable figure merging local and overall conspicuousness, the threshold value that have employed dense SIFT flow vector field criteria of quality evaluation and setting makes adaptive approach convert even convergent-divergent in time to by slit cropping, ensures that the edge distortion of the important area that traditional slit cropping easily occurs or not in image important area.Experimental result shows that the image after the method self-adaptation can while meeting display size, and ensure the display of image important area information completely, the metamorphopsic distortion of image drops to minimum.
Technical solution of the present invention is as follows:
A kind of image adaptive display packing, the perception of the method image content-based, in conjunction with slit cropping and even convergent-divergent, concrete steps are:
(1) set up visual attention location model: the respectively topical controls conspicuousness of each pixel and overall situation contrast conspicuousness in computed image, significantly figure and the overall situation are significantly schemed to obtain local; Remarkable for local figure and the remarkable figure of the overall situation is weighted combination, obtains final remarkable figure;
(2) initialization of finedraw: calculate energy cost figure as energygram with the remarkable figure obtained in step (1), select the finedraw that can remove according to energy cost figure, and be that the finedraw that in image, these can be removed sorts according to the low order high to energy cost of energy cost;
(3) measurement of slit cropping and image cropping distortion: carry out slit cropping according to the sequence in step (2), and utilize the distance between the intermediate images that generates in SIFT stream calculation original image and cutting process simultaneously;
(4) cut out termination: the distance between the intermediate images generated in original image and tailoring process exceedes threshold value θ, when namely deforming more than setting index, stop slit cropping; If the image miss the mark size after cutting out, then adopt even Zoom method to obtain the image of target size.
Especially, the specific implementation step of step (1) is:
A. adopt the change window of slip to calculate the topical controls value of the brightness of each pixel, texture and color: input picture gaussian pyramid is carried out multiple-stage filtering, calculate the topical controls figure of brightness, texture and color characteristic, take iterative interpolation summation algorithm, obtain the characteristic pattern on original image yardstick, the brightness obtained, texture and color characteristic figure normalization are also combined to the topical controls conspicuousness obtaining each pixel;
B. the overall situation contrast conspicuousness of each pixel in the overall situation contrast computed image based on color characteristic is adopted: using the difference of the difference of color characteristic as feature between two pixels, image is transformed into CIEL*a*b color space from rgb space, adopt the mode of image block, calculate current pixel block and the Euclidean distance around between other all block of pixels and and it can be used as the overall control value of this block of pixels central pixel point;
C. the overall situation is contrasted conspicuousness and detect foundation as main marking area, topical controls is significantly schemed be weighted combination with the remarkable figure of overall situation contrast, obtain final remarkable figure.
Especially, the specific implementation step of step (2) is:
A. the remarkable figure calculated according to step (1) calculates energy minimum cost figure;
B. energy minimum cost is looked for stitch according to energy minimum cost figure;
C. from significantly scheming and deleting energy minimum cost seam original RGB figure, remaining pixel is postponed successively to upper right side and is formed continuous print figure;
D. to the remarkable figure repeated execution of steps a to c that step c is formed, the location matrix that have recorded the finedraw that energy cost needs from high to low remove is obtained.
Especially, the specific implementation step of step (3) is:
A. the range equation between two images is defined: obtain the dense SIFT flow vector field between two width figure, calculate the mean value of all SIFT point distances that two width figure mate as the range equation between image, be used for the deflection of important information between evaluation objective picture and original image;
B. to utilize in step (2) and the finedraw sequence of cutting can carry out successive adjustment to picture size, remove 10 finedraws at every turn, and on the intermediate images at every turn obtained, calculate the distance of it and original image.
The present invention is by building a kind of marking area (i.e. image attention region) effectively extracting image based on local and the overall visual attention model contrasting conspicuousness, this attention model not only considers overall situation contrast but also combines topical controls conspicuousness, and the marking area therefore extracted is accurately complete; Slit cropping is carried out to image in the remarkable figure basis of this model extraction, and by the number of finedraw that the threshold restriction of setting can be cut out, produce geometric deformation distortion in slit cropping is by image before, is converted to even convergent-divergent.Image after the self-adaptation that the inventive method obtains can image significant concern area information complete image fault to be made to drop to again minimum.Adaptive approach of the present invention is that the lifting of the service quality of multimedia service in the integration of three networks provides important Technical Reference value.
Accompanying drawing explanation
Fig. 1 is principle framework figure of the present invention;
Fig. 2 is the structure process flow diagram of visual attention model in the present invention;
Fig. 3 is definite threshold θ lab diagram example;
Fig. 4 is comparing of the lab diagram that obtains with other adaptive approachs of the inventive method.
Embodiment
Present invention employs MSRAdatabase and RetargetMe two picture libraries, first several figure this method are carried out self-adaptation operation and determine optimal threshold θ; Then the target image that other adaptive approachs in the image after the self-adaptation of the present invention's generation and picture library (comprising direct shear (cropping), traditional finedraw shearing (seamcarving), evenly convergent-divergent (scaling), Non-uniformed mapping (WARP) etc.) obtain is compared.
Fig. 1 gives the frame diagram of the inventive method, by shown flow process, comprises following concrete steps:
(1) structure of visual attention model.
Fig. 2 gives the process flow diagram that this step visual attention model builds.As shown in Figure 2, in the present invention, the construction method of visual attention model is mainly divided into two parts: local conspicuousness calculates and overall conspicuousness calculates.
A. local conspicuousness calculates and adopts multiple dimensioned, to become the low-level features of window topical controls method.Its specific algorithm is as follows:
First, the change window slided is adopted to calculate the topical controls value of the brightness of each pixel, texture and color characteristic; When calculating the control value of each position pixel, this position is corresponding with the center pixel of window, calculates contrasting of the region of this point and surrounding window size, and the value of gained is as the topical controls value of this location point;
Brightness, texture, color are calculated respectively, obtains three width topical controls figure.
The brightness topical controls figure computing formula of (Weber-Fechner) rule is received as follows based on weber-Fick:
Wherein, I
cMthe brightness control value that (x, y) is pixel (x, y) place, c is constant,
with
be the brightness maxima in a jth window and mean value respectively, N '=(2k '+1) × { 1,2,3} represents the number of pixels in the change window of 3 different scales to (2k '+1) k ' ∈.
Computing formula based on the texture topical controls figure of gray variance is as follows:
HSI color space RGB image being transformed into view-based access control model perception carries out topical controls calculating, and method is as follows: first to two value of color Y at HSI color space
1=(H
1, S
1, I
1)
tand Y
2=(H
2, S
2, I
2)
t, definition color difference is:
Wherein Δ
i=| I
1-I
2|,
Therefore, color topical controls is calculated as follows formula:
In order to embody the topical controls of each point more accurately, the topical controls value of each pixel in 3 different scale windows is added the final control value as this point, and the final topical controls value of each pixel is calculated by following formula:
C_Map(x,y)=Σ
N′CM(x,y)
Wherein C_Map={I '
cM, T '
cM, C '
cMrepresent final brightness, texture and color characteristic topical controls figure, CM={I
cM, T
cM, C
cMrepresenting topical controls figure in single window, N '=(2k '+1) × { 1,2,3} represents the size of mutative scale moving window to (2k '+1) k ' ∈.
Then, in order to strengthen topical controls conspicuousness, the present invention adopts multi-scale method to calculate local conspicuousness, and input picture gaussian pyramid is carried out multiple-stage filtering, down-sampling obtains the image of original image on 6 different scales, wherein the corresponding input picture of the first yardstick.Along with sample stage, other increases, and the resolution of image reduces gradually.On every one-level yardstick, respectively according to the topical controls figure calculating brightness, texture and color characteristic.Each feature corresponding obtains the topical controls figure of three kinds of features under 6 width different scales respectively.Obtain 18 width topical controls figure altogether.Then a kind of iterative interpolation summation algorithm is taked, that is: upwards interpolation, summation step by step from the yardstick that resolution is minimum.Final at the highest yardstick of resolution, namely original image yardstick obtains the characteristic pattern of 3 width difference corresponding brightness, texture and colors.
Finally, because 3 width characteristic patterns are obtained by algorithms of different, the scope of acquired results is different, adopts maximal value normalization operator N (x) that 3 width characteristic patterns are normalized same scope, then merge and obtain last local and significantly scheme S_Local, method is as follows:
B. overall conspicuousness calculates and adopts the overall situation contrast based on color characteristic to obtain, using the difference of the difference of color characteristic as feature between measurement two pixels.In overall conspicuousness calculates, adopt the mode of image block, in CIE L*a*b color space, calculate the Euclidean distance of current pixel block and other all block of pixels of surrounding and as the overall control value of this block of pixels central pixel point.Overall situation contrast significantly figure S_Global is obtained by following formula:
S_Global(k)=Σ
jdis(p
k,p
j)
Wherein dis (p
k, p
j) represent the Euclidean distance of two block of pixels in CIE L*a*b color space.K is the pixel of current calculating, and S_Global (k) is larger, represents that this pixel overall situation conspicuousness is larger.
C., after obtaining local and the remarkable figure of the overall situation, topical controls is significantly schemed S_Local and overall situation contrast and significantly scheme S_Global and carry out combining the result ω that obtains as the weights of the remarkable model S_Global of the overall situation, obtain final remarkable figure S.Wherein ω
1and ω
2be respectively the weight coefficient that S_Local and S_Global combines, and meet Σ
iω
i=1.
ω=ω
1N(S_Local)+ω
2N(S_Global)
S=ω*S_Global
(2) initialization of finedraw: with the remarkable figure obtained in step (1) as energygram, calculate the finedraw can removed in image, i.e. the unessential information of image.
A. the remarkable figure calculated by step (1) is to calculate energy minimum cost figure M: for any one pixel (i in remarkable figure, j), its (i in energy minimum cost figure, j) value of position be in the remarkable map values of adjacent three pixels of the saliency value of the remarkable figure in this position and its upper row minimum one add with, that is: M (i, j)=S (i, j)+min (S (i-1, j-1), S (i-1, j), S (i-1, j+1)).Cost value for the first row of energy minimum cost figure is the value being directly set as remarkable figure.The vertical finedraw of cutting and the horizontal finedraw of cutting need their energygram of separate computations, for the method for the energygram of the vertical finedraw of cutting presented hereinbefore, should use the same method after only the remarkable figure of input picture need being carried out transposition and just can obtain the energygram that level cuts out.
B. the energy cost figure of image looks for optimum seam with regard to starting after obtaining: first find the pixel (i that the energy cost value of last row of image is minimum, j), preserve the position of this pixel, and then look for the point of minimum energy cost value in adjacent three pixels toward last row, be i.e. min (M (i-1, j-1), M (i-1, j), M (i-1, j+1)), this position is preserved.Repeat this process until first row, and the position of all pixels of this vertical lap seam found is saved in the position vector that a finedraw cuts out.
C., after optimum seam finds, the path forming this optimum seam is just deleted by from remarkable figure and original RGB figure, remaining pixel can towards right or upward the mobile continuous print figure that formed so that optimum seam is found in ensuing continuation.
D. get back to step a, repeat above step, just can obtain the location matrix that have recorded the finedraw that energy cost needs from low to high remove.
(3) measurement that finedraw is cut out and image cutting-out is out of shape.
A. the dense SIFT flow vector field of two width figure is calculated.Utilize SIFT feature to calculate vector field Q (p)=[u (p), v (p)] between two width figure, u (p) and v (p) is respectively horizontal component and the vertical component of p point vector field.The method calculating dense SIFT flow field Q is as follows:
Q when above formula calculation optimization objective function O (Q) is minimum.Wherein, S
iand S
jfor image I
iand I
jsIFT feature, σ is constraint factor, and d is smoothing factor.Q is respectively horizontal component and the vertical component of q point vector field at the 4 field set of pixels ε of p, u (q) and v (q).Section 1 tolerance I in above formula
iand I
jthe similarity of two width figure SIFT feature, Section 2 retrains SIFT vector field, and around last term priority match motion continuous print, adjacent pixels is to reach the object of smooth flow.Above formula solve exactly O (Q) minimum time flow field Q.By dense SIFT flow vector field Q, the range equation that we define between two images is as follows:
Wherein, N is original image vegetarian refreshments sum, N
sIFTfor the number of coupling SIFT point out.We carry out the deflection of important information between evaluation objective picture and original image with this equation, and the value of imgDist is larger, then two map distances are far away, and the more former figure of target image is out of shape more serious; The value of imgDist is less, then two map distances are nearer, and the more former figure of target image is out of shape fewer.
B. to utilize in step (2) and the finedraw sequence of cutting can carry out successive adjustment to picture size, remove 10 finedraws at every turn, and on the intermediate images at every turn obtained, calculate the distance of it and original image.
(4) termination is cut out.
A. relatively and judge whether the image distance that the i-th step calculates exceedes threshold value θ, if do not exceed threshold value, then continue to compare the image distance that (i+1) step calculates and whether exceed threshold value θ; Otherwise forward step b to;
B. stop comparing of image distance and threshold value θ in slit cropping, and continue use even convergent-divergent algorithm the intermediate images of (i-1) step is directly zoomed to final size.
We choose 20 width pictures in picture library MSRAdatabase and have carried out size change over operations and statistical study, the main information observing image after this method self-adaptation under different threshold value preserves situation, determine best threshold value, make the image of generation give prominence to pith, reduce again the geometric distortion of object.Fig. 3 is the picture that the embodiment threshold value selected chooses process, and wherein a is classified as original image, b row, c row, d row, e row be θ value are respectively 25,30, the later picture of 35,40 self-adaptations generated.We know that θ is larger, and the finedraw cut out is more, and the object of prospect is more outstanding, and this can find out from picture dog, flower, hot air balloon; But along with the increase of θ, finedraw cuts out increase, important objects just there will be the geometric distortion of profile, and this can find out from picture capsicum.Picture by great many of experiments and after considering self-adaptation is outstanding important objects both, reduces again the contour distortion of important objects, and we select setting threshold value θ to be 35.
We also choose the picture that in picture library RetargetMe, various adaptive approach generates, and the picture after the self-adaptation that obtains of the method that proposes of the present invention compares, and evaluation criterion is exactly the image distance equation with the definition of SIFT stream.Fig. 4 is some the good lab diagrams selected from numerous experiment, wherein a is classified as original image, b is classified as the image generated through direct shear (cropping), c is classified as and cuts out through finedraw the image that (seam carving) generate, d is classified as the image that even convergent-divergent (scaling) generates, e is classified as the image that Non-uniformed mapping (WARP) generates, the image that f generates after being classified as self-adaptation of the present invention.From this few width figure, we can find out that direct shear method lost the many important informations of image, finedraw is cut out and image outline can be caused to be out of shape, even convergent-divergent can make important objects in image produce deformation, Non-uniformed mapping makes image insignificant region and important area all produce certain distortion, and our method well highlights important area have compressed insignificant region and be out of shape less.The vector current obtained with dense SIFT flow vector field calculates the distance between two width images, and range formula is as follows:
Wherein u (p) and v (p) is respectively horizontal component and the vertical component of q point vector field, and N is original image vegetarian refreshments sum, N
sIFTfor the number of coupling SIFT point out.This range formula is just as the evaluation criterion of adaptive approach quality, and the distance calculated is less, then after self-adaptation, the important information deformation of the more former figure of image is less, and namely adaptive approach is more excellent; Vice versa.The form provided by table 1 represents the adaptive quality under the criteria of quality evaluation that picture that the 7 width pictures of random selecting in the picture having done from RetargetMe and tested and several adaptive approach generate defines at SIFT stream, the evaluation result being designated as black matrix is exactly that the inventive method obtains the highest result of adaptive quality relatively down at other adaptive approachs, although the self-adaptation picture that other this method obtain there is no the highest evaluation quality, the quality evaluation result obtained also is good.
Direct shear | Slit cropping | Non-uniformed mapping | The present invention | |
Ship | 119.9538 | 118.9071 | 118.9071 | 104.6988 |
Fairground | 148.8196 | 131.8392 | 128.7678 | 125.2993 |
House, canal | 106.6217 | 132.3944 | 132.7691 | 130.5573 |
Deck | 137.8378 | 153.2638 | 148.0140 | 136.1202 |
Cat | 80.5048 | 98.9550 | 92.1723 | 81.3691 |
Household | 80.1433 | 85.8651 | 87.3965 | 80.0126 |
Fish | 90.6655 | 144.9651 | 139.9793 | 124.6329 |
Table 1
Claims (4)
1. an image adaptive display packing, the perception of the method image content-based, in conjunction with slit cropping and even convergent-divergent, concrete steps are:
(1) set up visual attention location model: the respectively topical controls conspicuousness of each pixel and overall situation contrast conspicuousness in computed image, significantly figure and the overall situation are significantly schemed to obtain local; Remarkable for local figure and the remarkable figure of the overall situation is weighted combination, obtains final remarkable figure;
(2) initialization of finedraw: calculate energy cost figure as energygram with the remarkable figure obtained in step (1), select the finedraw that can remove according to energy cost figure, and be that the finedraw that in image, these can be removed sorts according to the low order high to energy cost of energy cost;
(3) measurement of slit cropping and image cropping distortion: carry out slit cropping according to the sequence in step (2), and utilize the distance between the intermediate images that generates in SIFT stream calculation original image and cutting process simultaneously;
(4) cut out termination: the distance between the intermediate images generated in original image and tailoring process exceedes threshold value θ, when namely deforming more than setting index, stop slit cropping; If the image miss the mark size after cutting out, then adopt even Zoom method to obtain the image of target size.
2. image adaptive display packing according to claim 1, is characterized in that: the specific implementation step of step (1) is:
A. adopt the change window of slip to calculate the topical controls value of the brightness of each pixel, texture and color: input picture gaussian pyramid is carried out multiple-stage filtering, calculate the topical controls figure of brightness, texture and color characteristic, take iterative interpolation summation algorithm, obtain the characteristic pattern on original image yardstick, the brightness obtained, texture and color characteristic figure normalization are also combined to the topical controls conspicuousness obtaining each pixel;
B. the overall situation contrast conspicuousness of each pixel in the overall situation contrast computed image based on color characteristic is adopted: using the difference of the difference of color characteristic as feature between two pixels, image is transformed into CIEL*a*b color space from rgb space, adopt the mode of image block, calculate current pixel block and the Euclidean distance around between other all block of pixels and and it can be used as the overall control value of this block of pixels central pixel point;
C. the overall situation is contrasted conspicuousness and detect foundation as main marking area, topical controls is significantly schemed be weighted combination with the remarkable figure of overall situation contrast, obtain final remarkable figure.
3. image adaptive display packing according to claim 1, is characterized in that: the specific implementation step of step (2) is:
A. the remarkable figure calculated according to step (1) calculates energy minimum cost figure;
B. energy minimum cost is looked for stitch according to energy minimum cost figure;
C. from significantly scheming and deleting energy minimum cost seam original RGB figure, remaining pixel is postponed successively to upper right side and is formed continuous print figure;
D. to the remarkable figure repeated execution of steps a to c that step c is formed, the location matrix that have recorded the finedraw that energy cost needs from high to low remove is obtained.
4. image adaptive display packing according to claim 1, is characterized in that: the specific implementation step of step (3) is:
A. the range equation between two images is defined: obtain the dense SIFT flow vector field between two width figure, calculate the mean value of all SIFT point distances that two width figure mate as the range equation between image, be used for the deflection of important information between evaluation objective picture and original image;
B. to utilize in step (2) and the finedraw sequence of cutting can carry out successive adjustment to picture size, remove 10 finedraws at every turn, and on the intermediate images at every turn obtained, calculate the distance of it and original image.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510227064.XA CN104835114A (en) | 2015-05-06 | 2015-05-06 | Image self-adaptive display method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510227064.XA CN104835114A (en) | 2015-05-06 | 2015-05-06 | Image self-adaptive display method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN104835114A true CN104835114A (en) | 2015-08-12 |
Family
ID=53812984
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510227064.XA Pending CN104835114A (en) | 2015-05-06 | 2015-05-06 | Image self-adaptive display method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104835114A (en) |
Cited By (26)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106971376A (en) * | 2017-04-20 | 2017-07-21 | 太原工业学院 | A kind of image-scaling method based on conspicuousness model |
CN107608751A (en) * | 2017-09-14 | 2018-01-19 | 光锐恒宇(北京)科技有限公司 | A kind of method and apparatus of display image in the application |
CN109615670A (en) * | 2018-12-05 | 2019-04-12 | 广西大学 | A kind of multi-operator image reorientation method based on compression domain perception of content |
CN110377773A (en) * | 2019-07-17 | 2019-10-25 | Oppo广东移动通信有限公司 | Image processing method, device, mobile terminal and storage medium |
CN110853068A (en) * | 2019-09-30 | 2020-02-28 | 华为技术有限公司 | Picture processing method and device, electronic equipment and readable storage medium |
US11403069B2 (en) | 2017-07-24 | 2022-08-02 | Tesla, Inc. | Accelerated mathematical engine |
US11409692B2 (en) | 2017-07-24 | 2022-08-09 | Tesla, Inc. | Vector computational unit |
CN114911551A (en) * | 2021-02-08 | 2022-08-16 | 花瓣云科技有限公司 | Display method and electronic equipment |
CN114928692A (en) * | 2021-02-12 | 2022-08-19 | 菲力尔商业***公司 | Image non-uniformity mitigation systems and methods |
US11487288B2 (en) | 2017-03-23 | 2022-11-01 | Tesla, Inc. | Data synthesis for autonomous control systems |
US11537811B2 (en) | 2018-12-04 | 2022-12-27 | Tesla, Inc. | Enhanced object detection for autonomous vehicles based on field view |
US11561791B2 (en) | 2018-02-01 | 2023-01-24 | Tesla, Inc. | Vector computational unit receiving data elements in parallel from a last row of a computational array |
US11562231B2 (en) | 2018-09-03 | 2023-01-24 | Tesla, Inc. | Neural networks for embedded devices |
US11567514B2 (en) | 2019-02-11 | 2023-01-31 | Tesla, Inc. | Autonomous and user controlled vehicle summon to a target |
US11610117B2 (en) | 2018-12-27 | 2023-03-21 | Tesla, Inc. | System and method for adapting a neural network model on a hardware platform |
US11636333B2 (en) | 2018-07-26 | 2023-04-25 | Tesla, Inc. | Optimizing neural network structures for embedded systems |
US11665108B2 (en) | 2018-10-25 | 2023-05-30 | Tesla, Inc. | QoS manager for system on a chip communications |
US11681649B2 (en) | 2017-07-24 | 2023-06-20 | Tesla, Inc. | Computational array microprocessor system using non-consecutive data formatting |
US11734562B2 (en) | 2018-06-20 | 2023-08-22 | Tesla, Inc. | Data pipeline and deep learning system for autonomous driving |
US11748620B2 (en) | 2019-02-01 | 2023-09-05 | Tesla, Inc. | Generating ground truth for machine learning from time series elements |
US11790664B2 (en) | 2019-02-19 | 2023-10-17 | Tesla, Inc. | Estimating object properties using visual image data |
US11816585B2 (en) | 2018-12-03 | 2023-11-14 | Tesla, Inc. | Machine learning models operating at different frequencies for autonomous vehicles |
US11841434B2 (en) | 2018-07-20 | 2023-12-12 | Tesla, Inc. | Annotation cross-labeling for autonomous control systems |
US11893393B2 (en) | 2017-07-24 | 2024-02-06 | Tesla, Inc. | Computational array microprocessor system with hardware arbiter managing memory requests |
US11893774B2 (en) | 2018-10-11 | 2024-02-06 | Tesla, Inc. | Systems and methods for training machine models with augmented data |
US12014553B2 (en) | 2019-02-01 | 2024-06-18 | Tesla, Inc. | Predicting three-dimensional features for autonomous driving |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102063258A (en) * | 2010-12-31 | 2011-05-18 | 山东大学 | Self-adaptive display method for image on display terminals with different sizes |
CN103955887A (en) * | 2014-04-11 | 2014-07-30 | 北京工业大学 | Content-based image self-adaptation method with combination of rapid seam carving and non-uniform mapping |
-
2015
- 2015-05-06 CN CN201510227064.XA patent/CN104835114A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102063258A (en) * | 2010-12-31 | 2011-05-18 | 山东大学 | Self-adaptive display method for image on display terminals with different sizes |
CN103955887A (en) * | 2014-04-11 | 2014-07-30 | 北京工业大学 | Content-based image self-adaptation method with combination of rapid seam carving and non-uniform mapping |
Non-Patent Citations (3)
Title |
---|
DAW-SEN HWANG 等: "CONTENT-AWARE IMAGE RESIZING USING PERCEPTUAL SEAM CARVING WITH HUMAN ATTENTION MODEL", 《PROCEEDINGS OF 2008 IEEE INTERNATIONAL CONFERENCE ON MULTIMEDIA AND EXPO》 * |
张静 等: "基于SIFT相似性度量的图像缩放", 《计算机工程》 * |
王金庭 等: "一种改进的Seam Carving图像缩放算法", 《红外技术》 * |
Cited By (31)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US12020476B2 (en) | 2017-03-23 | 2024-06-25 | Tesla, Inc. | Data synthesis for autonomous control systems |
US11487288B2 (en) | 2017-03-23 | 2022-11-01 | Tesla, Inc. | Data synthesis for autonomous control systems |
CN106971376A (en) * | 2017-04-20 | 2017-07-21 | 太原工业学院 | A kind of image-scaling method based on conspicuousness model |
US11681649B2 (en) | 2017-07-24 | 2023-06-20 | Tesla, Inc. | Computational array microprocessor system using non-consecutive data formatting |
US11403069B2 (en) | 2017-07-24 | 2022-08-02 | Tesla, Inc. | Accelerated mathematical engine |
US11409692B2 (en) | 2017-07-24 | 2022-08-09 | Tesla, Inc. | Vector computational unit |
US11893393B2 (en) | 2017-07-24 | 2024-02-06 | Tesla, Inc. | Computational array microprocessor system with hardware arbiter managing memory requests |
CN107608751A (en) * | 2017-09-14 | 2018-01-19 | 光锐恒宇(北京)科技有限公司 | A kind of method and apparatus of display image in the application |
US11561791B2 (en) | 2018-02-01 | 2023-01-24 | Tesla, Inc. | Vector computational unit receiving data elements in parallel from a last row of a computational array |
US11797304B2 (en) | 2018-02-01 | 2023-10-24 | Tesla, Inc. | Instruction set architecture for a vector computational unit |
US11734562B2 (en) | 2018-06-20 | 2023-08-22 | Tesla, Inc. | Data pipeline and deep learning system for autonomous driving |
US11841434B2 (en) | 2018-07-20 | 2023-12-12 | Tesla, Inc. | Annotation cross-labeling for autonomous control systems |
US11636333B2 (en) | 2018-07-26 | 2023-04-25 | Tesla, Inc. | Optimizing neural network structures for embedded systems |
US11983630B2 (en) | 2018-09-03 | 2024-05-14 | Tesla, Inc. | Neural networks for embedded devices |
US11562231B2 (en) | 2018-09-03 | 2023-01-24 | Tesla, Inc. | Neural networks for embedded devices |
US11893774B2 (en) | 2018-10-11 | 2024-02-06 | Tesla, Inc. | Systems and methods for training machine models with augmented data |
US11665108B2 (en) | 2018-10-25 | 2023-05-30 | Tesla, Inc. | QoS manager for system on a chip communications |
US11816585B2 (en) | 2018-12-03 | 2023-11-14 | Tesla, Inc. | Machine learning models operating at different frequencies for autonomous vehicles |
US11537811B2 (en) | 2018-12-04 | 2022-12-27 | Tesla, Inc. | Enhanced object detection for autonomous vehicles based on field view |
US11908171B2 (en) | 2018-12-04 | 2024-02-20 | Tesla, Inc. | Enhanced object detection for autonomous vehicles based on field view |
CN109615670A (en) * | 2018-12-05 | 2019-04-12 | 广西大学 | A kind of multi-operator image reorientation method based on compression domain perception of content |
US11610117B2 (en) | 2018-12-27 | 2023-03-21 | Tesla, Inc. | System and method for adapting a neural network model on a hardware platform |
US12014553B2 (en) | 2019-02-01 | 2024-06-18 | Tesla, Inc. | Predicting three-dimensional features for autonomous driving |
US11748620B2 (en) | 2019-02-01 | 2023-09-05 | Tesla, Inc. | Generating ground truth for machine learning from time series elements |
US11567514B2 (en) | 2019-02-11 | 2023-01-31 | Tesla, Inc. | Autonomous and user controlled vehicle summon to a target |
US11790664B2 (en) | 2019-02-19 | 2023-10-17 | Tesla, Inc. | Estimating object properties using visual image data |
CN110377773A (en) * | 2019-07-17 | 2019-10-25 | Oppo广东移动通信有限公司 | Image processing method, device, mobile terminal and storage medium |
CN110853068B (en) * | 2019-09-30 | 2022-06-17 | 荣耀终端有限公司 | Picture processing method and device, electronic equipment and readable storage medium |
CN110853068A (en) * | 2019-09-30 | 2020-02-28 | 华为技术有限公司 | Picture processing method and device, electronic equipment and readable storage medium |
CN114911551A (en) * | 2021-02-08 | 2022-08-16 | 花瓣云科技有限公司 | Display method and electronic equipment |
CN114928692A (en) * | 2021-02-12 | 2022-08-19 | 菲力尔商业***公司 | Image non-uniformity mitigation systems and methods |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104835114A (en) | Image self-adaptive display method | |
CN108596055B (en) | Airport target detection method of high-resolution remote sensing image under complex background | |
CN102708370B (en) | Method and device for extracting multi-view angle image foreground target | |
CN106909902B (en) | Remote sensing target detection method based on improved hierarchical significant model | |
CN101520894B (en) | Method for extracting significant object based on region significance | |
CN103049763B (en) | Context-constraint-based target identification method | |
CN106611420B (en) | The SAR image segmentation method constrained based on deconvolution network and sketch map direction | |
CN108596108B (en) | Aerial remote sensing image change detection method based on triple semantic relation learning | |
CN111027497B (en) | Weak and small target rapid detection method based on high-resolution optical remote sensing image | |
CN102402685B (en) | Method for segmenting three Markov field SAR image based on Gabor characteristic | |
CN110309842B (en) | Object detection method and device based on convolutional neural network | |
CN113408594B (en) | Remote sensing scene classification method based on attention network scale feature fusion | |
CN108376232A (en) | A kind of method and apparatus of automatic interpretation for remote sensing image | |
CN111401380B (en) | RGB-D image semantic segmentation method based on depth feature enhancement and edge optimization | |
CN102254326A (en) | Image segmentation method by using nucleus transmission | |
CN104392241A (en) | Mixed regression-based head pose estimation method | |
CN111640116B (en) | Aerial photography graph building segmentation method and device based on deep convolutional residual error network | |
DE102018003475A1 (en) | Form-based graphic search | |
CN106846322A (en) | Based on the SAR image segmentation method that curve wave filter and convolutional coding structure learn | |
CN104217440A (en) | Method for extracting built-up area from remote sensing image | |
CN109034258A (en) | Weakly supervised object detection method based on certain objects pixel gradient figure | |
CN111563408A (en) | High-resolution image landslide automatic detection method with multi-level perception characteristics and progressive self-learning | |
CN110751606A (en) | Foam image processing method and system based on neural network algorithm | |
CN113052121B (en) | Multi-level network map intelligent generation method based on remote sensing image | |
CN113963009A (en) | Local self-attention image processing method and model based on deformable blocks |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
EXSB | Decision made by sipo to initiate substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20150812 |