Modeling Image Sparsity in Compressive Sensing
Compressive sensing is an emerging technology which can recover a sparse signal vector of dimension N via a much smaller number of non-adaptive, linear measurements than N. It is stated that the K-sparse signal can be recovered exactly from ( log( / )) M K N K ?? measurements provided the measurement matrix satisfies the so-called restricted isometry property (RIP). However, for the compressible signal, such as the image, which is not K-sparse, how many measurements it requires to achieve an acceptable visual quality? In this paper, we study the relationship between the image complexity and the required measurements in compressive sensing. We propose a mathematical model based on image texture and edge density to estimate the number of needed measurements. The experimental results with a large number of natural images shows that, quite most reconstructed images using our pre-calculated number of measurements have good enough quality (PSNR > 32dB), which confirms our proposed complexity-based model well.
Shanzhen Lan Pin Xu Qi Zhang
Communication University of China, Beijing
国际会议
2011亚太信号与信息处理协会年度峰会(APSIPAASC 2011)
西安
英文
1-4
2011-10-18（万方平台首次上网日期，不代表论文的发表时间）