Image Zooming Based on Wavelet and Matrix-valued Diffusion Equation
一种基于小波和矩阵型扩散的图像放大

Based on wavelet multi-resolution analysis and matrix-valued diffusion equation,an image zooming model was presented.The original image was presented as the low pass filter of the wavelet analysis,and its high pass filters were zeros.The matrix-valued diffusion equation was used to the zoomed image and then wavelet analysis,the obtained high pass filter was used as the estimated high pass filter with the original image can be used to reconstruct.The new model adopts the idea of the property like the edge-enhancing diffusivity that the eigenvalues are designed so that the new diffusion coefficients have negative values in the gradient direction,when the gradient is large,to sharpen the edges and positive ones along the edge direction to enhance them; simultaneously a comparability criterion of the zoomed image and the original one were proposed.So the image was denoised and edge preserved as well as zoomed.Some numerical experiments illustrate that the proposed methods is a valid new image zoomed algorithm.