An image region of interest extraction method based on two-step clustering of eye movement data

Abstract

The invention provides an image region of interest extraction method based on two-step clustering of eye movement data, which comprises a pre-clustering stage and a clustering stage. The purpose of pre-clustering is to eliminate the noise points in the eye movement data, input the multi-human eye movement trajectory point data on the test picture, combine the trajectory point data of each two people to cluster, remove the noise points with the recognition standard of the point level and the class level, and retain the eye movement trajectory points belonging to the effective fixation process.The purpose of clustering is to extract ROI from the effective eye-movement trajectory points. With all the effective eye-movement trajectory points as input, the points are firstly clustered, in which each category represents a ROI, the center represents the location of the ROI, and the number of points within the category represents the amount of attention that the region attracts. The inventionanalyzes the multi-person trajectory points through two-step clustering, and the obtained region of interest extraction result is more consistent with the human visual attention mode, and has betterstability and anti-noise interference ability.

Description

technical field [0001] The invention relates to the field of eye movement data analysis, in particular to a method and system for extracting regions of interest based on two-step clustering of eye movement data. Background technique [0002] The eye tracker can be used to record the eye movement characteristics of the human eye when processing visual information, including eyeball position, fixation duration and fixation position and other related information. With the intelligent development of eye tracker hardware and related software, eye movement data is widely used in psychological Academic research fields such as science and medicine, and commercial fields such as advertising and web page optimization. Among them, one of the application hotspots of the eye tracker is to find the user's interest area for a certain visual target through the distribution characteristics of the user's eye movement data. Correspondingly, extracting image regions of interest based on eye mo...

AI Technical Summary

Problems solved by technology

However, when the target contains content of different shapes and randomly distributed, such as natural scene images, the manual labeling method cannot efficiently complete the task of extracting the region of interest, especially when the number of images is large, the manual labeling method needs to consume a lot of time. time and manpower

Images