Authors : Yong Sang Ryu, Jin Soo Kim, Young Soo Park and Sang Hun Lee
Abstract: This study proposes a superpixel algorithm which divides boundary of an object by increasing color contrast of original image using edge information. A bilateral filter was used to remove texture of an object's surface from original image. Then edge is detected by difference of value blurring using a Gaussian filter. The color contrast of original image is improved and the Simple Linear Iterative Clustering (SLIC) algorithm is applied. In the existing algorithms, boundaries are not clear due to small variations of color values, resulting in segmentation into same object and a non-uniform size and shape, depending on texture. However, in the proposed algorithm, edges with edge features are detected to improve color contrast of edge positions which are divided along boundaries and reduce result of unnecessary division according to texture. The proposed Superpixel algorithm can improve the response rate by increasing necessary division by texture and dividing more precisely according to boundary.
Yong Sang Ryu, Jin Soo Kim, Young Soo Park and Sang Hun Lee, 2018. A Study on a Proposed Superpixel Algorithm by Increasing Color Contrast of the Original Image Using Edge Information. Journal of Engineering and Applied Sciences, 13: 9451-9456.