基于哈达玛矩阵编码测量的自适应压缩成像

为了消除采样过程中的噪声干扰,进一步提高重构图像质量,针对数字微镜阵列(DMD)与桶探测器在测量过程中点对点采样产生的起伏噪声导致图像信噪比降低的问题,提出基于哈达玛矩阵编码测量的压缩采样成像方法。首先采用DMD分区控制方法,利用哈达玛编码测量,计算获得低分辨率的粗糙图像,接着在预测的重要小波系数所在区域,对同一尺度上的重要区域利用哈达玛矩阵进行投影,同时计算出这些区域的小波系数,最后通过小波逆变换获得重构图像。实验表明,在测量噪声为0.2倍的热噪声下,只需要10%的采样率,通过哈达玛编码测量,图像峰值信噪比从13.98dB最高提高到34.56dB,提高了20.58dB,成像质量明显改善,清晰度高。当存在较大的测量噪声时,该方法可以大幅提高图像的信噪比,尤其适用于微弱光信号条件下的高灵敏压缩采样成像。

Adaptive compressed sampling imaging based on hadamard matrix coding measurement

In order to eliminate the noise interference in the sampling process and to further improve the quality of reconstructed images, a compressed sampling imaging method based on Hadamard matrix encoding measurement is proposed to reduce the signal to noise ratio of the Digital Micro-mirror Device (DMD) and the bucket detector in the measurement process. First, using the DMD zoning control method, the rough image of low resolution is obtained by the Hadamard coding measurement. Then, the important region of the predicted wavelet coefficients is projected with the Hadamard matrix on the important region on the same scale, and the wavelet coefficients of these regions are calculated. Finally, the wavelet inverse transformation is used to get the reconstructed images. The experiment shows that under 0.2 times the thermal noise, only 10% of the sampling rate is needed. The peak signal to noise ratio(PSNR) of the image is increased from 13.98 dB to 34.56 dB by the Hadamard code, which is improved by 20.58 dB, the image quality is obviously improved and the definition is better. Under weak light signal, when there is a large measurement noise, the method can achieve high SNR with low sampling rate.