基于二维Tsallis熵改进PCNN图像分割.pdfVIP

第 38卷第 4期 东南大学学报 ( 自然科学版 ) V ol38 No4 2008年 7月 JOURNAL O F SOUTHEAST UN VI ERSITY ( Natural Scie ce Editio ) July 2008 T sallis PCNN 张煜东 吴乐南 (, 210096) : 为了改善图像分割的性能, 采用改进的脉冲耦合神经 络( PCNN) 进行分割, 通过对其 内部活动项进行空不变的单阈值化分割, 来达到对原图像空变阈值化分割效果. 另外分割准则 也作了修正, 通过计算图像二维直方图的T sallis熵, 得到二维 T sallis熵, 以此作为图像分割准 则. 最后, 修正了动态门限项的下降速度, 使得 PCNN 收敛更快. 实验证明二维 T sallis熵准则优 于最大 Sha o 熵准则与最小交叉熵准则, 且改进的 PCNN 模型比传统 PCNN 模型收敛更快. : 图像分割; 二维直方图; T sallis熵; 脉冲耦合神经 络 : TN91173 : A : 1001- 0505( 2008) 04-0579-06 Image segmentation based on 2D Tsallis entropy with improved pulse coupled neural networks Zha g Y udo g W u Le a ( Schoo l of I form atio Scie ce a d E g i eeri g, SoutheastU iversity, Na j i g 210096, Chi a) Abstract: I order to am eliorate trad itio al mi age segm e tatio , a mi proved pulse coup led eural etw ork ( PCNN) is i troduced. The i er fu ctio item of the PCNN is looked as a ew mi age, w hich is segm e ted w ith space-i varia t thresho ld to achieve the effect that orig i al mi age is segm e - ted w ith space-varia t thresho ld. M ea w h ile, T sallis e tropy is com bi ed w ith 2D h istogram to e - ge der 2D T sallis e tropy, w h ich is used for gu idi g mi age segme tatio . F i ally, the decrease ve- locity of dy am ic threshold is accelerated i order to m ake the PCNN co verge faster. Expermi e ts show that 2D T sallis e tropy perform s better tha maxmi um Sha o e tropy a dm i mi um cross e - tropy, a d the mi proved PCNN ca co verge m ore quickly tha co ve tio al PCNN. K ey words: mi age segme tatio ; tw o-dmi e sio al h istogram; Tsallis e tropy; pu lse coup led eural etw ork o . T