Detection of early infarct signs on nonenhanced CT is mandatory in patients with acute ischemic stroke. Loss of the gray-white matter interface at the lentiform nucleus or the insular ribbon has been an important early infact sign, which affects decisions on thrombolytic therapy. However, its detection is difficult, since early infarct sign is of subtle hypoattenuation. To improve the detectability of early infarct sign, image processing that could reduce local noise while preserving edges is desirable. To examine this issue, we devised an adaptive partial smoothing filter(APSF). Since the APSF markedly improves visibility of the normal gray-white matter interface, loss of the gray-white matter interface due to hypoattenuation could be more easily detected. The APSF was applied to clinical CT images in hyperacute stroke patients. Our preliminary results showed that the visibility and detectability of early infarct signs was much improved. To validate the usefulness of the proposed method, two commonly used smoothing filters were also employed for comparison. The results demonstrated the superiority of the APSF. Our proposed APSF can improve the visibility of the gray-white matter interface, thereby enhancing the detectability of early infarct signs.