Adaptive background suppression method based on visual receptive field

作者：Zhang Zhen; Chen Zhe; Lv Li; Wang Xin; Xu Lizhong^*

来源：Chinese Journal of Scientific Instrument, 2014, 35(1): 191-199.

• Background suppression
• Gaussians
• NIR Imaging
• Optical environment
• Visual receptive fields

In the application of river surface imaging velocimetry, the water flow tracers shown as dim and small targets are easily affected by complex background noises, such as shadows and reflections, which leads to large errors in displacement estimation. To solve this problem, firstly the distribution and statistics characteristics of target, background and noise in NIR river surface images are analyzed to build the mathematical model. Then, inspired by the biological phenomenon of lateral inhibition, an adaptive background suppression method is presented based on the visual receptive field difference of Gaussian (DOG) model. To achieve local optimization of enhancement, the model parameters are determined using the prior knowledge of the intensity distributions of targets and noises in the river surface images, and the constraint relation that the excitatory and inhibitory effects compensate for each other; and the local optimization enhancement effect is achieved. The experiment results show that, as a band-pass filter, the DOG model is superior to traditional spatial high-pass filter in the performance of target enhancement, background suppression and noise filtering. The images obtained with the proposed method not only have good visual effects, but also meet the requirement of sufficient signal-noise ratio (SNR) for the correlation operations in subsequent motion vector estimation.

• 出版日期2014
• 单位河海大学