Results |
||||||||||||||||||
In some cases it might be difficult to notice the subtle differences between the results. Here we allow to compare between the automatic multi-operator results and those of single operators and our user study. Simply hover the mouse over a thumbnail image to switch the retarget display to the corresponding result. The following have been used for abbreviation: N-mixed: N-operator mixed path; N-regular: N-operator regular path. Please refer to the paper for further details on the algorithm and the operators used. Note: javascript must be enabled for this page to work correctly. Also, please allow some time for the page to load, as the images are used in uncompressed format. |
||||||||||||||||||
|
||||||||||||||||||
|
||||||||||||||||||
Figure 9: An illustration of the dynamic programming table used to optimize the search for the best mixed path using two operators only - seam carving (SC) and scaling (SL). The colors in table (b) indicate the BDW distance of the best image in each step. The retargeted result is shown in (c) - this is the best result using a mixed path (i.e. the algorithm automatically determines the order of operators and how much each should contribute). The optimal operator sequence found is <-30SL,-30SC,-10SL,-20SC,-10SL,-10SC,-10SL,-20SC,-10SL>. For comparison, we show the results of using two regular paths (d) <-70SL,-80SC> and (e) <-80SC,-70SL>, and the optimal regular path (f) <-90SC,-60SL>. (g) uses scaling and (h) seam carving.
|
||||||||||||||||||
Figure 13: We find the optimal regular path by finding the minimal BDW score (the red dot in (b)) for a combination of two operators (seam carving and scaling in this case) to retarget an image. Compare the resulting image (c) to using just seam carving (d) or just scaling (e).
|