Comparing to Bidirectional Similarity (Bidir) [1]
and Seam Carving (SC) [2].
See Sec. 5.2 in the paper for more details.
Results in this page are retargeted to scales/aspect ratios in range x0.4 to x1.75
Knitting Pattern
input image
BiDir produces result that locally looks very good but lacks global structure, when retargeting to large scales, the overall knitting pattern is not preserved.
SC yields reasonable results for most scales and aspect ratios.
Our InGAN is able to maintains not only the local patterns, but also the global look of the knitting pattern.
Birds
input image
Capitol Building
input image
Fruits stand
input image
Carrots
input image
Farm house
input image
Plowing
input image
References
S. Avidan and A. Shamir. Seam carving for content-awareimage resizing.ACM Trans. Graph., 26(3), July 2007.
D. Simakov, Y. Caspi, E. Shechtman, and M. Irani. Sum-marizing visual data using bidirectional similarity. InTheIEEE Conference on Computer Vision and Pattern Recogni-tion (CVPR), 2008.