An Improved Object Detection Algorithm Based on the Hessian Matrix and Conformable Derivative

In this paper, a newfangled technique for edge detection that combines the Khalil conformable derivative and the Hessian matrix is developed and experimentally validated. The following main aspects are considered: (i) to attenuate image noise a Gaussian kernel inspired by the Khalil conformable derivative was developed. (ii) The spatial derivatives of the image gray level are calculated via the Khalil derivative, which is suitable for maintaining object contours and texture information even in low-contrast and resolution images. (iii) The conformable Hessian matrix is obtained from these image derivatives, generating continuous, thicker, and brighter edges. Our operator is compared with some existing techniques. Simulation results on different test images confirm a greater robustness to noise by our operator as well as a better visual quality of the edge maps obtained. This statement is validated through a comparative analysis based on peak signal-to-noise ratio and edge-strength-similarity-based image quality. In addition, it is applied in medical image processing and analysis tasks such as mammogram images, computed tomography scans, and magnetic resonance imaging for identification tasks of middle cerebral artery aneurysms, calcifications and breast cancer, proliferative Diabetic Retinopathy, and cerebral arteriovenous malformations. According to our findings, the conformable operator allows better visual identification of the structure of these conditions, improving the accuracy of clinical diagnosis and its subsequent monitoring.