Offline signature verification (SigVer) is regarded as one of the most challenging tasks in all biometrics. In the case of signature biometrics, the skilled forgeries are created with a high level of precision and expertise. Hence, extracting intricate features from signature images plays a significant role in discriminating between genuine and forgery images. This study aims to explore the possibility of bringing out the importance of feature reduction and feature selection mechanisms for efficient offline SigVer. The Histogram of Oriented Gradients (HOG) and Mutual Information (MI) feature selection criteria are considered for the experimentation purpose. Firstly, HOG features are obtained from the signature images, and reduction in dimensionality is accomplished through the use of MI, an effective filter-based feature selection technique. Finally, the authenticity of the test signature is decided using the SVM classifier. The proposed model is simple but effective in terms of reducing computation complexity. Extensive experimentation was conducted on the CEDAR dataset and achieved an accuracy of 97%, which outperforms the state-of-the-art result. The results obtained indicate the effectiveness and efficacy of the proposed model.