In today’s era of medical image analysis, computer-aided detection has emerged as a revolutionary approach for automated detection of disease, particularly in the case of limited-experienced endoscopists. The detection at an earlier stage is necessary to overcome other disorders. This paper implements a Gaussian blur for removing degradation in the image as a preprocessing step. For finding a particular region of interest, the Sobel operator is applied to the denoised image. The proposed novel approach extracts the texture and frequency details using the gray level co-occurrence matrix and three-level discrete wavelet transform. The classification into normal and abnormal images is achieved with a linear support vector machine, attaining an accuracy of 85.07%. The proposed approach attains sensitivity and specificity of 88.88% and 77.27%, respectively. The results are compared with other techniques like KNN, PNN, and DWT with nonlinear features. The high sensitivity achieved insights into cases of correctly identifying the abnormal images. The accuracy achieved is not 100% but it can be increased further with an increasing dataset. For early detection and treating the disease on time, this automated process can prove boon for clinical studies.