梯度先验模型耦合改进复扩散的图像放大

Image enlargement based on gradient prior model coupled improved complex diffusion

为提高图像中弱边缘和纹理细节的放大效果,提出一种梯度先验模型耦合改进复扩散的单幅图像分辨率增强算法。利用改进的复扩散和冲击滤波器对图像边缘进行增强并通过梯度约束模型对弱边缘进行增强,减少图像放大引起的混叠效应和吉普斯效应,通过变分模型实现两者的耦合,具有良好的放大效果。梯度先验模型通过放大图像和边缘插值图像梯度逼近实现,复扩散模型采用自适应扩散参数耦合冲击滤波器实现对边缘的增强。和其它算法进行仿真实验比较,比较结果表明,该算法在强边缘、弱边缘和细节方面能够取得较好的分辨率重建效果。

To improve the effects of the weak edge and texture in the zoomed image, the single image super-resolution model was proposed using gradient prior model coupled improved complex diffusion. Through the coupling of variation model of improved complex diffusion and gradient prior model, the improved complex diffusion enhanced strong edges better and gradient prior model enhanced the weak edge and texture details better. The proposed model reduced Gibbs effects and aliasing and had better performance. Through approximating gradient of the interpolated image, the high resolution image gradient was constrained using gradient prior. Adaptive complex diffusion coupled shock filter enhanced the edge. Results of simulation show that the proposed model has better enlargement effects on weak edge, strong edge and details.