Zhao H.-W.,Jilin University |
Zhao H.-W.,Key Laboratory of Symbolic Computation and Knowledge Engineering of the Ministry of Education |
Wang Z.,Jilin University |
Liu P.-P.,Jilin University |
And 2 more authors.
Guangxue Jingmi Gongcheng/Optics and Precision Engineering
A new image resizing algorithm was proposed based on a color space and an energy updating mechanism. First, the concept of color space was introduced into image resizing algorithm and the pixel's energy function was re-built based on the gradient of gray space and color space. Three kinds of color energy functions were tested and the energy function in HSI space was finally chosen because of its excellent performance. Additionally, since image resizing might badly corrupt the contours of main areas in an original image and might further affect the understanding of the image, an energy updating mechanism was proposed. The energy updating mechanism could increase the neighborhood energy values where a pixel had been deleted. By doing so, it could decrease the probability that seams cross this area again, thus the contour of this area could be protected well. In experiments, the degree of Scale Invariant Feature Transform (SIFT) matching between original image and resized image with different λ values was calculated. The best performance shows that the proportion between the color and gray energy functions was 7:3. By comparing the proposed method with other two state of arts algorithms, it indicates that the shape of main area will be better preserved with the proposed algorithm without introducing any more complexity. Source