Abstract Achieving the objective of identifying epileptic seizure activities automatically using electroencephalographic (EEG) signals is of great significance in the treatment of epilepsy. To realize this goal, a hybrid approach to analyze the time–frequency (t–f) image of EEG signals is employed in this study. In the proposed approach, the EEG signals are transformed into a t–f image using short-time Fourier transform and the t–f images are further decomposed into various component images by applying coiflet wavelet transformation. The texture descriptor, namely the local binary pattern, gray-level cooccurrence matrix, and local tetra pattern methods, are employed to compute the features from the wavelet-filtered images. The extracted features are fed into four different classifiers for t–f image classification. Experimental evaluation performed on two publicly available EEG datasets suggests that the proposed method is a proficient and powerful method for more accurate and earlier detection of epilepsy.