To address the issues of low efficiency and limited precision in wear detection for mechanical scribing tools, a machine vision-based inspection methodology was proposed to achieve cost-effective, high-efficiency wear monitoring with micron-level measurement accuracy. Firstly, the image of the cutting tool tip was obtained by an industrial camera, and then image cropping and grayscale processing were carried out on it. Subsequently, three methods, namely median filtering, Gaussian filtering and mean filtering, were adopted to filter and compare the grayscale images respectively. The filtered images were sharened and the edge features were optimized by selecting the best ones. Finally, the edge of the scratch tool image was extracted based on the Canny sub-pixel edge detection method. Results indicate that the main wear areas of the scribing tool are the primary scribing edge and flank face, with wear lengths of 361.1 μm and 43.7 μm respectively, while wear depth values gradually decrease to 0 μm along the direction away from the tool tip, with average wear depth values of 3.1 μm and 2.9 μm in these two areas. The maximum wear depth is concentrated at the tool tip with a value of 11 μm.