基于机器学习的低分辨率图像增强和细节匹配方法研究

当前低分辨率图像增强和细节匹配方法具有细节易丢失、边缘模糊、无法适应图像平移、旋转等变化的弊端,导致图像增强与细节匹配性能低下。为此,提出一种新的基于机器学习的低分辨率图像增强和细节匹配方法,通过建立一个间隔最大的超平面获取最小二乘支持向量机分类器。在待处理低分辨率图像中选择一块图像,将图像的每个3×3邻域像素看作一个训练样本,通过最小二乘支持向量机法对其进行训练,输出增强像素点。通过复数小波对图像特征进行描述,利用最小二乘支持向量机获取最优判定准则函数,输出最优匹配的目标子图像。实验结果表明,所提方法有很高的峰值信噪比、边缘保持指数和等效视数,很低的归一化均方误差、均值和方差,整体性能优。

Low resolution image enhancement based on machine learning research and details matching method

At present, the low resolution image enhancement and detail matching method is easy to be lost, blurred edge, can not adapt to the change of image translation, rotation and other defects, resulting in image enhancement and the performance of low matching performance. To this end, a new low resolution image enhancement and detail matching method based on machine learning is proposed. In the low resolution image to be processed, select an image, and the image of each 3 x 3 pixels as a training sample, the least squares support vector machine method to train the image, the output enhancement pixel. The image features are described by complex wavelet, and the least square support vector machine is used to obtain the optimal decision criterion function. The experimental results show that the proposed method has a high peak signal to noise ratio, the edge preserving index and the equivalent apparent number, and the low normalized mean square error, mean and variance, and the overall performance of the proposed method are excellent.