In order to ensure printing quality of text content in raster image, we usually need to adjust the thickness of characters in the pre-press. This paper presents a sub-pixel adjustment method, based on analysis of characters’ neighborhood pixel information, so that we can control the scope and direction of the adjustment. In our method, Image is considered as a combined result of a foreground character layer and a background image layer. By performing the steps of separating characters from background, calculating ink density of each pixel of characters, adjusting each pixel’s ink density by analysis of neighborhood stroke shape, we achieve a result so fine that could simulate the effects performed by vector methods. The method can be adapted to global or any certain direction (e.g. vertical, horizontal, diagonal, etc) thickness adjustment. Experimental results show that the method can achieve excellent results and is proved to be very practical.