Objective: Focal cortical dysplasias (FCDs) are a common cause of drug-resistant focal epilepsy and are often difficult to visually identify on MRI. Automated FCD detection has proven challenging, at least in part due to difficulty in accounting for variability of FCD lesions as well as of normal cortex. Here, we implemented a normative modeling approach based on a novel set of local imaging features to account for expected normal and lesional variability. Methods: A standardized epilepsy imaging protocol consisting of T1, T2 and FLAIR MR images was collected in 15 patients with radiologically or histologically diagnosed FCDs and 30 healthy volunteers (HVs). We computed a set of generic 3D multiscale local image filters across multiple MR contrasts, sampled onto the gray-white junction surface. We created a latent representation of normal cortical variability in HVs with a known probability distribution that allowed for outlier detection and estimates of similarity. Following local normalization, we implemented an automated approach to FCD detection, identifying outliers based on projection onto the mean FCD vector. Results: Our features could be used to predict features commonly measured in other FCD detection approaches. In our initial normative model, FCDs are outliers, but only to a similar extent as some normal cortical regions. FCDs have similar underlying features to some of these regions. Following local normalization, FCDs become more easily distinguishable. An outlier detection procedure based on projection onto the average FCD feature vector achieved a sensitivity of 87% and specificity of 70%. Significance: We present a novel set of generic local imaging features that appear to be useful in characterizing normal cortex as well as FCD lesions. Our normative modeling approach allows for identification of outliers and similarity. Together, they represent a novel approach to automated FCD detection and may also be of use in other imaging applications.