Abstract: Aiming at the problem of the weak ability of the 3-D transform to sparsely represent the true-image data in Block Matching 3D(BM3D) algorithm, an improved algorithm of image denoising by sparse BM3D is proposed. This paper uses shape-adaptive image patches(neighborhoods) replacing square windows image block. Shape-adaptive image patches by 3-D transform processing are PCA transformed. Experimental results show that the proposed method can efficiently denoise the Gaussian noise and improve the PSNR and SSIM of image, especially in preserving image details and introducing very few artifacts.

Key words: shape-adaptive image block, Principal Component Analysis(PCA), 3-D transform processing, sparsity, image denoising

摘要： 三维块匹配法中3-D变换真实信号的稀疏表达能力较弱，针对该问题，提出一种关于图像去噪的三维块匹配(BM3D)改进算法。采用形状自适应的图像块(邻域)代替BM3D算法中的平方窗图像块，对3-D变换处理的形状自适应图像块进行PCA变换。实验结果证明，该算法能够有效去除图像的高斯噪声，提高图像的峰值性噪比和结构相似度，且在保持图像的边缘等细节信息方面性能较好，图像视觉效果有明显改善。

关键词: 形状自适应图像块, 主成分分析, 三维变换处理, 稀疏性, 图像去噪

CLC Number: 

• TP391.41