基于学习的图像智能适配显示技术

为适应不同的显示分辨率,出现了各式各样的图像适配显示(IR)的方法.提出了基于图像列的一种快速适配显示方法.在处理过程中,首先,计算一个原始图像的重要性图;其次,根据图像每列的重要性程度为其分配一个比例因子,对不同图像而言,应对比例因子设置不同的上限才可以得到较好的结果;最后,提出通过机器学习方法计算出不同图像的上限,从而可以高效率地得到理想的结果.根据每一列的比例因子采用像素融合的方式处理图片得到目标分辨率.本方法是基于列实现的,其复杂度低、便于计算;设置每列系数的上限控制了图像重要部分的宽度,从而减少了不连贯,处理结果更为自然.

Learning based intelligent image retargeting technique

There has been a wide range of image retargeting (IR) approaches, in order to solve the problem of adapting images to different display resolutions. A fast image retargeting method was proposed, which was based on image columns. Firstly, the method would calculate a saliency map of the original image. Secondly, a group of scaling factors were generated for image pixel fusion, which was used to get the result image of the target image size. Each image column corresponded to its scaling factor. For different images, an adaptive upper bound was obtained by machine learning, for scaling factor assignment. This upper bound was set to limit the column width and can reduce image distortion. The experiment results prove that this adaptive upper bound results in a better performance. Moreover, this method has a low complexity, thus it calculates fast, as it is based on image columns instead of pixels.