Examination of road condition, especially pavement crack, needs to be done regularly. With regular checks, road conditions can be maintained to keep it comfortable for road users. Currently, the examination was carried out manually where the officers observe all pavement road and then take notes and mark the existence of cracks. This conventional method takes a long time, labour intensive and low consistency due to the human factor. To overcome the problem, this research proposes the use of digital image processing technique to detect the existence of cracked road surface. The detection technique is developed using Gaussian Pyramid method to create multiscale images and extraction of histogram and black-white area features. Linear discriminant analysis is then used to classify the crack and non-crack images. Based on experiment results, the method produces the accuracy of 92.8571% with 1.50 seconds of computation time per image. In conclusion, the proposed method successfully increase the accuracy of cracks detection on pavement surface