Abstract: The acoustic logging images are quite blurry with a lot of noise, which makes the automatic detection of reflectors difficult. It is time-consuming and labor-consuming to rely on expert recognition. Therefore, a completely automatic detection process is proposed. Each pixel in the logging images was solely split into one color cluster through the Gaussian mixture model to build multi-channel sub-images. The sub-images including reflectors were combined together. The coarse noise reduction was performed based on local connectivity, and the fine noise reduction was performed based on the number of pixels in the connected area. The accurate pixel-level detection of the reflector area was completed. The entire process was fully automated. Experiments were performed on the acoustic logging images used in oil-field development. This method achieved accurate pixel-level detection of the reflector area, which greatly improved development efficiency.