The weak fault features of wheel bearing are difficult to accurately detect because of the interference of strong background noise. Aiming at the problem, this paper presents a novel method named self adaptive improved Laplacian of Gaussian (ILoG) operator to detect the weak fault features of wheel bearing. The ILoG operator filter has excellent ability to detect the sudden change of signals, which is applied to detect the fault impulse characteristics in bearing fault signals. In addition, water cycle algorithm (WCA) with good optimization characteristic is used to search for the influencing parameters of ILoG operator in order to achieve the best filtering results. The envelope demodulation method is further used to analyze the best filtering signals of the optimized ILoG operator and extract weak fault features. The proposed method is applied to analyze wheel bearings with outer race and inner race fault, and the results show that this method can detect the weak fault characteristic frequencies of bearings effectively. The filtering effect is better than the wavelet threshold denoising and multi-scale morphological difference filter methods.