基于信息冗余的小波红外图像去噪算法

摘要

红外图像具有图像灰度集中、对比度低等特点,因而红外图像增强是红外图像应用必不缺少的部分,随之而来的是图像噪声的放大,为了进一步提高红外图像质量,需对增强后图像去噪.现有众多去噪方法中,极少同时兼顾算法效果及可实现性.提出了一种基于信息冗余的小波去噪算法,此算法在离散小波变化(DWT)过程中分别以不同的下采样方式获取多组合有相似冗余信息的小波系数,再利用噪声估计对小波系数进行非线性变换,抑制高频噪声并保留细节,然后利用变换后小波系数重构(IDWT)多副图像,利用含相似冗余信息的多副图像加权进一步去除高频噪声,获取最终去噪图像.此算法已在单片FPGA中进行实现,利用ALTERA CYCLONIII芯片实现后的处理帧频达到50 fps,满足实时性要求.%Infrared images always have low contrast and narrow gray level distribution. Infrared image enhancement is requisite for further application, which usually leads to the amplification of noise. In order to improve the quality of infrared images, it is essential to de-noise the enhanced image. Existing de-noising methods rarely take into account both efficiency and feasibility. In this paper, a de-noising algorithm based on redundant digital wavelet transformation (DWT) is proposed. The algorithm obtains several group of wavelet coefficients that contain similarly redundant information by different down-sampling in the processing of DWT. A non-linear transformation on the wavelet coefficients is used to retain the details and repress the high frequency noise based on the result of noise estimation. Then the reconstruction of processed wavelet coefficients (IDWT) generates a set of mid-images which contain similarly redundant information. Last a final de-noised image is acquired by weighted adding of the mid-images that further remove the high frequency noise. This algorithm has been realized in a single FPGA implementation. The processing frame rate on the ALTERA CYCLONIII chip is 50 f/s,which meets the real-time requirements.