Studies have reported the fractal dimensional increment of glass and glass-ceramic fracture surfaces. The objective of this study was to determine the relationship between fracture toughness and fractal dimensional increment of two types of dental glass-ceramics with different volume fraction of crystals and different fracture surface roughness. Fifteen bar-shaped specimens were prepared from lithium disilicate (LDS) by sectioning the blocks (IPS e.max CAD, Ivoclar Vivadent). Ten beams were prepared by mixing and layering the nanofluorapatite (NFA) glass-ceramic powder (IPS e.max Ceram, Ivoclar Vivadent). One face of each of the specimens was indented using a Knoop diamond at 25 N (LDS) or 10 N (NFA) followed by loading in deionized water in 4-point, or 3-point flexure, respectively, until failure. Fracture surfaces were analyzed using scanning electron microscope (SEM). Fracture toughness (Kc) was calculated using the surface crack in flexure (SCF) technique (ASTM C1421). Epoxy replicas of the fracture surfaces were scanned using the atomic force microscope (AFM) followed by noise filtering. The FRACTALS software was used to determine the fractal dimensional increment (D*) by the Minkowski cover algorithm. In addition, surface roughness was computed from the AFM scans using the Gwyddion software. The median (25%, 75% quartiles) fracture toughness of LDS and NFA bars were 1.62 (1.59, 1.69) MPam1/2 and 0.68 (0.66, 0.74) MPam1/2, respectively. The median fractal dimension (D) value (25%, 75% quartiles) before and after noise filtering for LDS were 2.16 (2.15, 2.17) and 2.14 (2.14, 2.15), respectively, and for NFA were 2.29 (2.21, 2.38) and 2.17 (2.17, 2.18), respectively. The NFA follows the regression model (Y=1.99 X) between Kc and sqrt. D* for glasses as previous studies, but the LDS deviated from the model. The median (25%, 75% quartiles) surface roughness before and after noise filtering for LDS were 139 (119, 188) nm and 137 (118, 187) nm and for NFA were 7 (6, 15) nm and 7 (6, 15) nm, respectively. Noise filtering successfully eliminated noise from the AFM scans of the material with smooth fracture surfaces (NFA), resulting in a decrease in measured fractal dimension. The LDS data deviated from the model because of toughening due to crack bridging. Fractal analysis with noise filtering can be used to estimate the fracture toughness of dental glass-ceramics that do not exhibit crack bridging.