A fingerprint recognition system aims to identify individuals. The main challenge in fingerprint recognition is the low quality of the images. This low quality can be attributed to various factors, including oily or dry skin types and the type of fingerprint scanner used. Therefore, efforts are made to improve the fingerprint image quality, as it is a crucial factor in determining the accuracy of fingerprint recognition results. To make dirty fingerprint images more interpretable by both humans and machines, quality improvement is necessary by minimizing the dirty areas. This research aims to enhance the quality of dirty fingerprint images using the Gabor filter method and the learning vector quantization method for testing. The testing is conducted using four images for training in each class. There are a total of 10 classes in this testing, with each class consisting of 10 images. In the testing phase, with 75 dirty fingerprint images, an accuracy rate of 33.3333% is achieved.