Application And Research Of Sketch Target Detection Based On Visual Saliency

The efficient distribution of attention and the formation of visual cognitive process can be implemented by the human visual system through the physiological mechanism of visual saliency.A lot of visual information can be effectively dealt with through the mechanism of the visual saliency model.When process and analyze the information of image and video,significant area became a hot research topic.Significant area is the key area can be able to attract human visual attention,which is the visual attention can be focused on the region of interest of the image in a short period of time.In this paper,we study the content mainly includes:1)A new saliency detection algorithm based on plastic feature is proposed based on traditional saliency algorithm.Firstly,the algorithm utilizes the gray inconsistent operator as a means of local processing,depicting non uniform of local texture,which makes the brightness of the most salient pixels is increased;Secondly,using the improved FT algorithm to establish a new global quantitative method,to make the area increase significantly;Thirdly,in order to filter out the influence of isolated area significantly,the algorithm proposes a space weighted expression to conduct the linear processing on significant map,improving the contrast of overall background area.Simulation experiment demonstrates that the proposed new algorithm is more accurate than classical algorithms,at the same time in the quantitative indexes such as precision rate and recall rate,the new one has strong advantage.2)An improved SIFT feature extraction algorithm is proposed based on traditional SIFT algorithm.This algorithm combined with visual area detection,that is the detection of key point is conducted in the significant area obtained by the new significant algorithm,greatly reduced the number of extraction of feature points.The extracted feature points are in accordance with human visual perception,which makeensure the effectiveness of feature extraction points.Then conducting feature match through the improved SIFT algorithm.Finally conduct the simulation experiment of the affine invariant and scale invariance of the traditional SIFT algorithm and improved algorithm,the results show that the improved algorithm not only matching accuracy is higher than traditional SIFT algorithm,and the matching speed is faster.3)Combine the new significant detection algorithm and the improved SIFT algorithm,this paper applies visual significance to sketch target retrieval.The sketch target is interest region,and it is consistent with human visual attention mechanism.This paper uses the salient map to conduct the sketch retrieval,filter out the background region which is more conducive to sketch retrieval.