Research of Anti - Noise Image Salient Region Extraction Method
Bing Xu, Shi-Yi Xie, LI Zhen-de · Journal of Engineering Science and Technology Review · 2014
The existing image salient region extraction technology is mostly suitable for processing noise-free images, and there is a lack of studies on the impact of noise on images.In this study the adaptive kernel function was employed in image salient region detection.The salient property of a region was determined by the dissimilarities between the pixels of the image region and its surroundings.The dissimilarity was measured as a decreasing function associated with adaptive kernel regression.The proposed algorithm used multi-scale fusion method to obtain the salient region of the whole image.As adaptive kernel function has strong anti-noise characteristics, the proposed algorithm was characterized with the same robustness.A numerical simulation experiment was conducted on salient region extraction of images with noise and without noise.A comparison between this study's results and two existing salient region extraction methods revealed that the proposed method in this study was superior in its extraction accuracy of image salient regions and could reduce interference of image noise.