Method for extracting image attention by image based multi-characteristic integration

Abstract

The invention discloses a method for extracting image attention by image based multi-characteristic integration, comprising the following specific steps: (1) input an image and extract multiple characteristics of each domain respectively; (2) carry out integration of multiple characteristics and calculate domain attention; (3) convert the domain attention into attention of each pixel point k in the image by proximity degree of the pixel value of each pixel point in the image and mean value of each domain, and extract image attention. In the method, the integration of multiple characteristics is adopted, the domain attention is calculated, the position concerned by people eyes can be accurately located, and the whole domain and peripheral profile of a concerned object can be accurately highlighted, the visual sense requirement of people eyes is met, and the method has the characteristics of high accuracy and good real-time performance; and the method has good application prospect in the field such as machine vision, object division and target tracing and the like.

Description

technical field [0001] The invention relates to the technical fields of computer vision and image processing, in particular to a method for extracting image attention based on multi-feature fusion of images. Background technique [0002] With the advent of the information age and the rapid development of multimedia technology, the amount of information transmission is increasing, including not only text information and voice information, but also a large amount of image information and video information. People do not pay equal attention to all the content in the image, and people pay the most attention to the useful information. Extracting image attention and finding out the position and object that human eyes pay attention to in the image is an important problem in the field of image analysis, pattern recognition and computer vision, and it is also a difficult problem. As in November 1998, L.Itti et al. published the article "A Saliency-Based Visual Attention Model for Fa...

AI Technical Summary

Problems solved by technology

As in November 1998, L.Itti et al. published the article "A Saliency-Based Visual Attention Model for Fast Scene Analysis" in the 20th issue of the IEEE "Graph Analysis and Machine Intelligence" journal. This paper introduces the visual attention model, which is mainly used to predict the focus of the human eye, but with the above visual attention model, only the position of the most concerned object can be found but the specific shape of the object cannot be found

In 2007, Liu Tie and others published the article "Learning to detect a single salient object" on pages 1 to 8 of the IEEE "CVPR Proceedings". Local, regional and global characteristics of prominent objects, and use the conditional random field model to detect salient objects. Although the salient objects can be highlighted, the outline of the salient objects is blurred, which does not conform to the perception of the human eye.

Images