Background and Objective: COVID-19 outbreak was declared as a pandemic on 11th March 2020. The rapid spread of this highly infectious virus has distinguished it from other classes of viral and respiratory diseases. The reverse transcription-polymerase chain reaction (RT-PCR) test is most commonly used for the qualitative assessment of the presence of SARS-CoV-2. Due to the high false-negative rate of RT-PCR tests reported worldwide, chest x-ray imaging has proved to be a feasible alternative for the detection of COVID-19. The COV-ELM classifier aims to classify COVID-19 cases from the chest x-ray images using extreme learning machine (ELM). The choice of ELM in this work is based on the fact that ELM significantly shortens the training time with the least interventions required to tune the networks as compared to other neural networks. Methods: The proposed work is experimented on the COVID-19 chest x-ray (CXR) image data collected from three publicly available sources. The image data is preprocessed and local patterns are extracted by exploiting the frequency and texture regions to generate a feature pool. This pool of features is provided as an input to the ELM and a 10-fold cross-validation method is employed to evaluate the proposed model. Results: The proposed method achieved a macro average of f1-score is 0.95 in a three-class classification scenario. The overall sensitivity of the COV-ELM classifier is ${0.94 \pm 0.03}$ at 95% confidence interval. Conclusions: The COV-ELM outperforms other competitive machine learning algorithms in a multi-class classification scenario. The results of COV-ELM are quite promising which increases its suitability to be applied to bigger and more diverse datasets.