纹理对比度导向的图像自适应压缩感知测量方法

发布时间：2018-02-26 04:22

本文关键词： 图像压缩感知自适应测量纹理对比度自适应全局重建　出处：《铁道学报》2017年06期 　论文类型：期刊论文

【摘要】：压缩感知以部分随机变换代替全变换,仍可确保图像的高精度复原,可用于铁路系统中无线监控终端实现低复杂图像编码。传统图像压缩感知以相同的测量率实施分块测量,而由于分块稀疏度不同,常导致重建图像具有块效应,无法确保良好的率失真性能。为了解决该问题,本文提出利用图像纹理特征引导图像压缩感知,在感知端实施自适应测量。利用像素八连通区域内的最大梯度度量各像素的纹理变化程度,生成纹理特征图,利用纹理特征图计算各块纹理对比度,并以此为依据自适应设定各块的测量率,以块纹理对比度加权图像重建模型的目的函数,集中优化纹理细节的区域。实验结果表明,与由块方差、边缘特征主导的自适应测量方法相比,本文所提算法可确保较好的重建图像主观视觉质量,且率失真性能优于传统压缩感知重建算法。
[Abstract]:Compression perception can replace full transform with partial random transformation, which can ensure high precision restoration of image, and can be used to realize low complex image coding at wireless monitoring terminal in railway system. Traditional image compression perception performs block measurement at the same measurement rate. However, due to the difference of block sparsity, the reconstructed image often has block effect, which can not guarantee good rate-distortion performance. In order to solve this problem, this paper proposes to use texture features to guide image compression perception. Using the maximum gradient in the eight connected regions of pixels to measure the texture variation degree of each pixel, the texture feature map is generated, and the texture contrast of each block is calculated by using the texture feature graph. Based on this, the measurement rate of each block is set adaptively, and the object function of the image reconstruction model weighted by block texture contrast is used to optimize the region of texture details. The experimental results show that, compared with the variance of block, Compared with the edge feature dominated adaptive measurement method, the proposed algorithm can ensure better subjective visual quality of reconstructed image, and the rate-distortion performance is better than that of the traditional compression perceptual reconstruction algorithm.
【作者单位】：信阳师范学院计算机与信息技术学院;南京邮电大学江苏省图像处理与图像通信重点实验室;北京交通大学计算机与信息技术学院;
【基金】：国家自然科学基金(61501393)
【分类号】：TP391.41

【相似文献】