Modified LACIF filtering in background disjoint noise

详细信息    查看全文

• 作者：Walid Ben Tara^a ; Pascuala Garcí ; a-Martí ; nez^b ; Henri H. Arsenault^a ; ^{arseno@phy.ulaval.ca"" rel=""nofollow}
• 关键词：Change detection ; Disjoint noise ; Intensity invariant pattern recognition ; Nonlinear correlations ; Vector spaces
• 刊名：Optics Communications
• 出版年：2011
• 出版时间：1 September 2011
• 年：2011
• 卷：284
• 期：19
• 页码：4430-4434
• 全文大小：1350 K

文摘

This work deals with pattern recognition methods based on correlations for images in the presence of noise. We propose a modification of the nonlinear Locally Adaptive Contrast Invariant Filter (LACIF) that yields correlation peaks that are invariant to linear intensity changes of the target but that has some limitations in the presence low variance nonoverlapping background noise. The modification of the filter implies a normalization by a global variance of several distributions. The estimation of the variance distributions is done locally by means of correlations. Experimental results as well as comparisons with the classical matched filter and the common LACIF are given.