By Topic

A new hybrid PCNN for multi-objects image segmentation

Sign In

Cookies must be enabled to login.After enabling cookies , please use refresh or reload or ctrl+f5 on the browser for the login options.

The purchase and pricing options are temporarily unavailable. Please try again later.
6 Author(s)
Zhenbo Li ; Coll. of Inf. & Electr. Eng., China Agric. Univ., Beijing, China ; Yu Jiang ; Jun Yue ; Jingjing Fang
more authors

Many image based applications such as multi-object tracking were nagged by the problem of robust multi-objects image segmentation. In this paper, we propose a new hybrid Pulse Coupled Neural Network (PCNN) method for multi-object segmentation. Firstly, we use saliency detection methods, Graph-based visual saliency (GBVS) and Spectrum Residual (SR) to find more accurate object region (R1) and more number of object regions (R2) separately. Then an improved PCNN is used to work out the multi-objects with R1 and R2. The statistical result of R1 is selected as an adaptive generator threshold of PCNN and a selection standard of segmentation result. R2 determines the correct object number in the image. Experiments of images selected from BSD and VOC and two full image datasets (MSRC v2 and Weizmann) prove that our method can get more right object quantity and more accurate object region than GBVS-PCNN[1] and adaptive PCNN[2].

Published in:

Signal & Information Processing Association Annual Summit and Conference (APSIPA ASC), 2012 Asia-Pacific

Date of Conference:

3-6 Dec. 2012