基于稀疏表示的图像模糊度评价方法

Image Fuzzy Degree Assessment Method Based on Sparse Representation

根据自然图像稀疏表示基函数类似Vl区简单细胞感受野的反应特性,以及系数反映神经元响应程度的特性,提出基于稀疏表示的图像模糊度评价方法。将输入图像分成互不重叠的块,采用训练好的词典对各块进行稀疏分解,并计算每块的关注度系数。将每块稀疏系数的p范数与关注度系数的加权和作为模糊度评价的依据。实验结果表明,该算法计算的模糊度相对于图像的模糊程度是单调的,具有较好的抗噪性,符合人眼视觉系统特性。

Based on the facts that natural images can be sparsely coded, and their basis functions are similar to the particular shapes of V1 simple-cell receptive fields and the sparse coefficients correspond to the response properties of visual neurons for the fixed patterns, a novel image fuzzy degree assessment method based on sparse representation is proposed. The flow chart of the proposed method is to divide the input image into no overlapped patches. The coefficient vectors and the visual attention weights of each patch are computed, and the fuzzy degree is represented by the linear superposition of the p-norm of sparse coefficient vectors and visual attention weights of each patch. Experimental results show that the proposed method is monotonic, robust to additive noises, and also consistent with human visual system.