Your SlideShare is downloading. ×
0
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
study Seam Carving For Content Aware Image Resizing
Upcoming SlideShare
Loading in...5
×

Thanks for flagging this SlideShare!

Oops! An error has occurred.

×
Saving this for later? Get the SlideShare app to save on your phone or tablet. Read anywhere, anytime – even offline.
Text the download link to your phone
Standard text messaging rates apply

study Seam Carving For Content Aware Image Resizing

4,427

Published on

study

study

Published in: Technology
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total Views
4,427
On Slideshare
0
From Embeds
0
Number of Embeds
0
Actions
Shares
0
Downloads
92
Comments
0
Likes
0
Embeds 0
No embeds

Report content
Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
No notes for slide
  • Shai Aidan 以色列人MERL(Mitsubishi Electric Research Labs)沒想到以色列人投效到日本研究中心
  • Transcript

    • 1. Seam Carving for Content-Aware Image Resizing<br />Shai Aidan (Mitsubishi Electric Research Labs)<br />Ariel Shamir (The Interdisciplinary Center & MERL)<br />ACM SIGGRAPH 2007<br />
    • 2. Resize<br />Seam carving & insertion <br />
    • 3. Abstract<br />Seams are optimal 8-connected paths of pixels cross the image<br />Carving out or inserting seams to achieve content-aware resizing <br />
    • 4. Outline<br />Introduction<br />Background<br />Seam-carving operator<br />Discrete image resizing<br />Multi-size images<br />Limitations<br />Conclusions and future work<br />
    • 5. Introduction<br />
    • 6. Motivation<br />HTML can support dynamic changes of page layout and text. Why can not an image deform to fit different layout automatically ?<br />iGoogle<br />How about aspect ratio of an image , such as fitting photo into PDA or phone cells ?<br />Solution ?<br />Resize – content independent<br />Crop – remove pixels from the image periphery only<br />
    • 7. Basic Idea of Seam-Carving<br />Use energy function to define the importanceof pixels<br />Define seam-carving image operator<br />Image reduction<br />Carving out seams - the connected low energy pixels crossing the image <br />Preserving the image structure<br />Image enlarging<br />Insert seams on low energy area<br />The order of seam insertion ensures a balance between the original image content and the artificially inserted pixels<br />
    • 8. Application<br />Discrete image resizing<br />Aspect Ration Change, Image Retarget, Image Enlarging, Content Amplification, Seam Carving in gradient domain, Object Removal<br />Multi-size images<br />An image can continuously change their size in a content-aware manner<br />Storing the order of seam removal and insertion<br />
    • 9. background<br />
    • 10. Image Retarget<br />Seek to change the size of the image while maintaining the important features <br />Face detector<br />An automatic thumbnail creation [Suh03] <br />ROI<br />Fisheye-View warp [Liu and Gleicher 05, 06]<br />Visual saliency [] <br />[Suh 03]<br />[Selur 04, decompose image to foreground obj and background<br />origin<br />
    • 11. Saliency map<br />[Itti IEEE99]<br />Simulate neuroscience of human visual system<br />Pyramid tech. to compute 3 feature maps, color, intensity and orientation<br />[Suh 03], an automatic thumbnail creation, based on either a saliency map or the output of a face detector<br />[Chen 03], adapting most important region of images to mobile devices. <br />
    • 12. [Liu 03], suggesting to trade time for space. Given a collection of regions of interest, they construct an optimal path through these regions and display them serially.<br />
    • 13. [Santella et al. 06] use eye tracking, in addition to composition rules to crop images intelligently. <br />
    • 14. ROI (Region-Of-Interest)<br />Such a method was proposed by [Liu and Gleicher 05, 06] for image and video retargeting. For image retargeting they find ROI and construct a novel Fisheye-View warp that essentially applies a piecewise linear scaling function in each dimension to the image. This way the ROI is maintained while the rest of the image is warped. The retargeting can be done in interactive rates, once the ROI is found, so the user can control the desired size of the image by moving a slider. In their video retargeting work they use a combination of image and saliency maps to find the ROI. Then they use a combination of cropping, virtual pan and shot cuts to retarget the video frames.<br />
    • 15. Feature-aware warping<br />The first solution to the general problem of warping an image into an arbitrary shape while preserving user-specified features was recently proposed by [Gal et al. 06]. <br />The feature-aware warping is achieved by a particular formulation of the Laplacian editing technique, suited to accommodate similarity constraints on parts of the domain. <br />Since local constraints are propagated by the global optimization process, not all the constraints can always be satisfied at once<br />
    • 16. Seam<br />Perfect seams to combine parts of a set of photo into a single composite picture [Agarwala et al. 04] <br />Drag-and-Drop Pasting that extends the Poisson Image Editing to computer an optimal boundary (seam) between the source picture and target images [Jia et al. 06]<br />AutoCollage, a program that automatically creates a collage image from a collection of images. [Rother et al. 06]<br />Simultaneously solve matting and compositing. They allow the user to scale the size of the foreground object and paste it back on the original background. [Wang , Cohen 06]<br />evaluated several cost functions for seamless image stitching and concluded that minimizing an L1error norm between the gradients of the stitched image and the gradients of the input images performed well in general [Zomet et al. 05] <br />
    • 17. Sear Optimal Seam<br />Dijkstra’s shortest path algorithm [98]<br />Dynamic programming [Efros 01]<br />Graph cuts [Kwatra 03]<br />
    • 18. Seam-carving operator<br />
    • 19. Strategies of Image Reduction<br />Original<br />e1 energy<br />Optimal <br />global remove the lowest energy pixels<br />Pixel<br />remove the least energy in each row<br />
    • 20. Strategies of Image Reduction<br />Original<br />e1 energy<br />Column<br />removing columns with minimal energy<br />Crop<br />find a sub-win with the highest energy<br />
    • 21. Vertical Seam<br />
    • 22. Horizontal Seam<br />
    • 23. Optimal Seam Search<br />
    • 24. Optimal Seam Search<br />S<br />Dynamic Programming<br />G<br />
    • 25. e1 energy<br />
    • 26. Image Energy Preservation<br />The average energy of all pixels during resizing<br />
    • 27. Energy Functions<br />L1 and L2-norm of the gradient, saliency measure [Itti 99]<br />
    • 28. Histogram of Gradient (HoG)<br /><ul><li>Histogram of Gradient (HoG) [Dalal and Triggs 95]</li></ul>Dividing the image window into cells<br />For each cell accumulating a local 1-D histogram of gradient directions<br />Normalize cells by the measure of local histogram energy over larger blocks<br />The average gradient image<br />Weighted R-HOG descriptor<br />R-HOG descriptor<br />
    • 29. Energy Functions<br />Histogram of Gradient (HoG) [Dalal and Triggs 95]<br />max(HoG(I(x,y)) makes sure the seams run parallel to the edge of objects and not cross them<br />
    • 30. Energy Functions<br />Entropy <br />Compute the entropy over a 9 x 9 window and add it to e1<br />eEntropy(x,y) =<br />+ e1 (x,y)<br />
    • 31. Energy Functions<br />Segmentation and L1<br />Image segmentation [Christoudias 02] <br />Apply e1 on the results<br />
    • 32. No single e function performs well across all images<br />Similar range for resizing<br />e1 or eHoG works well<br />
    • 33. Discrete image resizing <br />Aspect Ratio Change, Retargeting with Optimal Seams-Order, Image Enlarging, Content Amplification,<br />
    • 34. Aspect Ratio Change<br />Carving-out /insert seams<br />Original<br />Original<br />Original<br />1D aspect ratio changing<br />
    • 35. 2D aspect ratio changing<br />Optimal Seams-Order Search<br />+<br />DynamicProgramming<br />= <br />+<br />min<br />
    • 36. Retargeting with Optimal Seams-Order<br />h-first<br />alternate<br />v-first<br />Transport map<br />Original<br />optimal<br />
    • 37. Image Enlarging<br />Find first k seams for removal<br />Duplicate them in order to arrive at I(-k)<br />origin<br />I(t): smaller image after t seam-carving<br />I(-k): enlarged image after k seam insertion<br />t<br />I(-1)<br />I(-k)<br />I(t)<br />I(-k)<br />enlarged image<br />insert seams in order of removal<br />
    • 38. Image Enlarging (>50%)<br />origin<br />Break into several steps<br />Each step does not enlarge the size of image more than a fraction <br />
    • 39. Content Amplification<br />Amplified<br />Original<br />
    • 40. Seam Carving in the Gradient Domain<br />Seam + Poisson Reconstruction [Perez 03]<br />Compute e function<br />Work on the gradient domain<br />Remove seams from the x and y derivatives of the original image<br />Use Poisson Reconstruction<br />original<br />retarget<br />retarget in <br />Gradient Domain<br />
    • 41. Object Removal<br />Mark the removing target<br />Remove seams until all the marked pixels are gone<br />* Employ seam insertion to maintain the original size<br />
    • 42. Object Removal<br />Origin<br />
    • 43. Multi-size images<br />Store the pre-computed representation that encodes, for each pixel in V/H map<br />The index of the seam that removed it<br />The negative index of the seam that inserted it<br />Blue (first seam)  Red (last seam)<br />origin<br />V(i,j)=t : pixel (i,j) removed by t-th vertical seam<br />H(i,j)=t : pixel (i,j) removed by t-th horizontal seam<br />
    • 44. Limitations<br />Seam-Carving does not work well on all images<br /> Ex: face<br />Origin<br />Crop<br />Scale<br />Constraint the face<br />Face the flower<br />Bottom up feature detection<br />
    • 45. Limitations<br />The amount of content<br />Too density, no “less” important area<br />The layout of the image content<br />origin<br />origin<br />
    • 46. Conclusions<br />Present a content-aware resizing using the seam-carving image operator<br />Seams are the optimal paths on a single image<br />Carve-out seams<br />Insert seams<br />Application of seam-carving operator<br />Aspect ratio change, image retargeting, content amplification, object removal<br />Multi-size images that support continuous resizingin real-time<br />
    • 47. Future Work<br />Video resizing<br />Combination of scaling and seam-carving<br />Define more robust multi-size image<br />Better solution to combine horizontal and vertical seams in multi-size image<br />
    • 48. END<br />

    ×