The research presented in this thesis focused on the development of a prediction model for friction performance of asphalt pavements quantified as skid resistance, accounting for the nonstandard texture parameters. Pavement friction results from a complex interplay of many influencing parameters that can be grouped in four distinct groups: surface roughness properties, driving properties, vehicle tire properties and environmental influences. Surface roughness properties were selected as the key influencing factor in this research. There are two specific texture roughness scales relevant for pavement's friction performance: micro-texture and macro-texture. Current standardized practice enables the determination of macro-texture indicators, commonly related to friction performance measured on the roads. Despite the effort to establish a relationship between the standard texture indicators and friction performance, there still exists no unique model which would provide a reliable and unambiguous prediction of friction from the traditionally determined texture roughness properties. To investigate the relationship between pavement friction performance and surface roughness on both relevant texture scales, an alternative method based on remote sensing technology was developed in this thesis. The method utilized a digital camera for the acquisition of multiple pavement surface images from a close range, further used for the creation of a 3D digital surface model. The method was called Close-Range Orthogonal Photogrammetry - CROP method. Created 3D digital surface models enabled the analysis of multiple roughness parameters on micro- and macro-texture levels. The CROP method was optimized for the data acquisition procedure, photographic equipment used and procedure for digital surface model processing and analysis. The accuracy of CROP method was verified by performance comparison to a benchmark technology for 3D digital model creation – a high precision 3D laser scanner. Selected non-standard texture parameters were used as predictors in the development of a friction prediction model, performed in regression analysis framework. Friction performance was quantified by skid resistance measurements of the analysed surfaces, performed by a stationary low-speed measurement device. Four different regression-based models were established and compared for the model performance assessment, accounting for the model predictive strength evaluated by coefficient of determination values and selected error metric. The optimal model was defined by partial least squares regression, with two non-standard texture parameters selected as the most influential for the prediction of pavement surface friction performance. In comparison to the performance of simple linear regression model accounting for a single traditional texture indicator Mean Profile Depth, the model developed in the thesis obtained better performance for the prediction of skid resistance.