Saliency Detection Via Boundary Dissimilarity

The prominence of a target(ie,an object,a person,a pixel,etc.)is the state or quality that is prominent relative to its neighbors.Significant testing is considered a critical attention mechanism to promote learning and survival by focusing the organism on its limited perceptual and cognitive resources in the most relevant subset of available sensory data.The significance usually comes from the contrast between the project and its neighborhood,such as the red dot around the white point,the flicker message indicator of the answering machine,or the large noise in other quiet environments.Explicit detection is usually studied in the context of a visual system,but a similar mechanism operates in other sensory systems.What is significant This definition can be solved by training: for example,for a human subject,a particular letter can be highlighted by training.When attention is deployed by a significant stimulus,it is considered to be bottom-up,memoryless and reactive.Attention mechanisms can also be guided by top-down,memory-related,or expected mechanisms,such as moving forward or looking sideways before crossing the street.Humans and other animals are difficult to pay attention to multiple projects at the same time,so they are faced with the challenge of continually integrating and prioritizing different bottom-up and top-down effects.The saliency algorithm has become a practical tool for the use of CV field.At present,all the significant detection algorithms can be divided into two categories: top-down detection algorithm and bottom-up detection algorithm,the current mainstream is still bottom-up detection algorithm.In this paper,we propose a novel bottom-up approach for salient region detection based on boundary dissimilarity.This method is easy for implementation without any loss in quality.We obtain the saliency maps as follows: First,we decompose the image into numerous compact and visual homogeneous superpixels.Second,we calculate the saliency value of a superpixel by its dissimilarity to all the boundary superpixels,and we emphasize the effect of the boundary superpixel which is most nearest to this superpixel.At last,we would optimize our results by optimal formulas to obtain the ultimate saliency maps.Finally,a large number of comparative experiments are to prove the effectiveness of the algorithm and state-of-the-art results.