This paper describes a new approach for reducing ringing artifacts in spectroscopic imaging (or chemical-shift imaging). The new method is aimed at reducing the signal contamination in areas of interest, due to ringing, that often arises from adjacent areas with high intensity, such as subcutaneous fat in the case of 1H spectroscopy and skeletal muscle in case of the 31P spectroscopy. The method is based on the fact that the high-intensity signals which are the major source of contamination can be captured with one or few acquisitions and the metabolite resonances in areas of interest require considerably more averaging in the data acquisition due to low signal-to-noise ratio. In order to reduce ringing and maintain adequate SNR for the metabolites, the data points in high k space are measured with one or few excitations while those in the low k space are sampled with many excitations. By virtue of this sampling scheme, the data contain more noise in the high-spatial-frequency region and direct Fourier transformation reconstruction generates a very noisy image. Therefore, a new reconstruction technique is developed for the reconstruction. This approach has been studied by simulations and its practical utility is demonstrated by an in vivo proton CSI experiment. Technical details and simulation and experimental results are presented. From these results, it is concluded that the technique is robust and can be used for in vivo applications.