<p>Aiming at the problem that the traditional single algorithm cannot realize the accurate identification of Si₃N₄ bearing roller nanodefect images with low background contrast. A method for detecting nanodefects of Si₃N₄ bearing rollers based on a multiscale fuzzy connectivity active contour coupling segmentation algorithm is proposed. Combining the fuzzy features of nanodefect images, the mean filter matrix equation is designed to achieve image denoising and smoothing. Based on the fuzzy connection relationship segmentation, the initial contour of nanodefects is obtained. Analyzing the gradual change of defect and background in nanodefect images, the homomorphic filter function equation is defined to realize contrast enhancement of nanodefect images. The active contour model is built, and the initial contour evolves and iterates until it fits with the image defect contour. The characteristic image of nanodefects in Si₃N₄ bearing rollers is identified. After the noise reduction and enhancement processes, the average SSIM (structural similarity index measure) of Si₃N₄ bearing roller nanodefect images is 93.54%. The average PSNR (peak signal-to-noise ratio) is 33.83 dB. And the recognition accuracy of different types of defects in nanofeature images is not less than 93%.</p>