Face recognition is one of the interesting technologies recently and the study of face recognition was typically performed on images with various facial characteristics. Somehow, censored faces haven’t been directly studied yet. Therefore, this research conducted an experimental study of face recognition in 4 types of censorship on images: Gaussian Blur, Mosaic Blur, Gaussian Blur with Censor Bar, and Mosaic Blur with Censor Bar, by using Convolutional Neural Network (CNN) and Local Binary Patterns Histogram (LBPH), which were concluded from the previous research to be the 2 best methods for face recognition. Regarding the experimental results, the face recognition by CNN algorithm provided a significantly higher accuracy rate than LBPH. LBPH could not detect the face images having a censor bar, which made the faces unlike human ones since LBPH takes the Haar Cascade classifier in face detection by identifying regions of image where human faces are likely to be present. Furthermore, LBPH tended to recognize faces censored by Mosaic Blur with a lower accuracy rate compared to Gaussian Blur. Conversely, CNN was able to detect faces censored by all 4 types, with a tendency to recognize faces censored with Mosaic Blur at a higher accuracy rate than those censored with Gaussian Blur. In the case of the combination of burring and censor bar, CNN could recognize faces censored with Mosaic Blur*Censor Bar at a higher accuracy rate than Gaussian Blur*Censor Bar. In terms of censorship, the combination of burring and censor bar, Gaussian Blur*Censor Bar and Mosaic Blur*Censor Bar, provided the lowest accuracy rate of face recognition, which was significantly lower than Gaussian Blur or Mosaic Blur only.