Microwave open-ended waveguide scanning has been proved to be a promising NDT technique for imaging of woven carbon fiber reinforced plastic (CFRP) with impact damages. However, it uses a conventional C-scan technique, which is excessively time-consuming and therefore it is not practical for defect detection in a large area. Moreover, without proper feature extraction, the appearance of woven texture within the impact damage area opposes to the localization of the impact. This work proposes a novel framework for fast impact damage localization by means of discrete sparse measurement. Also, the localization accuracy of sparse measurement is improved by mitigating effect of woven texture using principal component analysis (PCA) for feature extraction. The performance of the proposed methods is evaluated by sets of incremental measurements. The performance of the proposed system is and evaluated by localizing the impact damage of 100 non-uniform sparse sampling sequences. Results show that our proposed technique can accurately localize the impact damage with number of samples less than 10% of full measurements, giving approximately 10 times faster than that of the conventional C-scan.