小目标识别的小波阈值去噪方法

De-noising algorithm of wavelet threshold for small target detection

为改善小目标识别的滤噪效果并提高其信噪比,构造了新的阈值函数并采用局部方差估计法来计算阈值对小目标进行去噪处理。对小波分解层次中各高频子带选取不同的阈值,其中大于阈值的小波系数采用改进的双曲线函数作为阈值函数,小于阈值的小波系数采用指数函数与对数函数相互组合的方式作为阈值函数。对采用的阈值函数进行了理论推导,并与软、硬阈值法进行了实验对比。计算机仿真结果表明:经本文阈值法处理后,信噪比相对于含噪图像提高了70.8%,而软、硬阈值法分别提高了49.8%和59.7%。光学实验进一步证实:该方法能更有效地提高信噪比,增强联合变换相关器对于小目标的识别能力。

In order to obtain better de-noising effects and higher Signal-to-noise Ratios（SNRs） for recognizing small targets,the local variance estimation method is adopted to calculate the threshold.Different thresholds are selected for all the high-frequency sub-bands in wavelet decomposition levels.The improved hyperbolic function is used as the threshold function for wavelet coefficients more than the thresholds and exponential and logarithmic functions are combined as the threshold function for wavelet coefficients less than the thresholds.The adopted threshold function is derived theoretically and compared experimentally with those of soft and hard threshold methods.Computer simulation results show that the SNR is improved by 70.8% with the threshold method adopted in this paper,while they are improved by 49.8% and 59.7% respectively by using soft and hard threshold methods.By optical experiments,it is further proved that the method can improve SNRs and enhance the recognition ability of small targets with Joint Transform Correlator（JTC） more effectively.