%0 Journal Article %T Blind Deconvolution for Ultrasound Sequences Using a Noninverse Greedy Algorithm %A Liviu-Teodor Chira %A Corneliu Rusu %A Clovis Tauber %A Jean-Marc Girault %J International Journal of Biomedical Imaging %D 2013 %I Hindawi Publishing Corporation %R 10.1155/2013/496067 %X The blind deconvolution of ultrasound sequences in medical ultrasound technique is still a major problem despite the efforts made. This paper presents a blind noninverse deconvolution algorithm to eliminate the blurring effect, using the envelope of the acquired radio-frequency sequences and a priori Laplacian distribution for deconvolved signal. The algorithm is executed in two steps. Firstly, the point spread function is automatically estimated from the measured data. Secondly, the data are reconstructed in a nonblind way using proposed algorithm. The algorithm is a nonlinear blind deconvolution which works as a greedy algorithm. The results on simulated signals and real images are compared with different state of the art methods deconvolution. Our method shows good results for scatters detection, speckle noise suppression, and execution time. 1. Introduction Medical ultrasound imaging is considered to be one of the edge technologies in noninvasive diagnose procedures. Despite its great advantages, as cost-benefit, accessibility, portability, and safety, it has a weak resolution. This is the result of the attenuations, refractions, nonlinearities, frequency selection, or probe properties [1]. As a result, important efforts were made in the direction of image quality improvement. Signal processing methods offer a reasonable approach for resolution improvement. From this point of view the most important methods for reconstruction are superresolution and deconvolution. If superresolution methods seem to be impractical, the deconvolution ones are more practical [2]. Supperresolution is a complex problem because of the difficulties in aproximation of reconstruction operators (e.g., motion, degradation, and subsampling operators) and the use of multiple frames which puzzles also the implementation. This was conducted on the proposition of multiple deconvolution approaches for ultrasound imaging, like methods used in system identification or Bayesian statistics based ones [3]. From these algorithms, the methods based on Bayesian approach, especially maximum a Posteriori (MAP) seem to offer the most interesting results [4¨C10]. In these methods the point spread function (PSF) is estimated and then the information is reconstructed in a nonblind way using a priori information about tissue reflectivity function. As the PSF estimation is an important problem that is complex, a lot of methods were advanced to propose an acceptable solution. Primary studies have considered a measured radio-frequency (RF) PSF [4, 11]. However, the use of only one RF PSF to deconvolve %U http://www.hindawi.com/journals/ijbi/2013/496067/