数字图像的固定基稀疏表示方法

Fixed-base sparse representation for digital images

提出了一种采用小波与轮廓波混合的固定基数字图像描述和变换方法,该变换在较粗糙的尺度上采用小波变换,在细节尺度上采用轮廓波变换,既能够克服可分离小波在图像描述方面高频子带方向分辨率较低的缺陷,又能够降低轮廓波变换的计算复杂度。实验结果表明,本方法在具有良好的数字图像逼近能力的同时,还具有很好的消噪结果。

This paper proposed a fixed base sparse representation by mixing wavelet transform and contourlet transform for digital images. The transform performed wavelet transform at coarser scales and contourlet transform at detail scales, thus could overcome the image representation limitations of separable wavelets due to its limitation for directional information and reduced the computation complexity of contourlet transform. Experimental results show that the mixed transform can approximate digital images efficiently, furthermore, the performance for image denoising good is very good.