Canny edge detection model in mri image segmentation using optimized parameter tuning method

Abstract

Image segmentation is a crucial stage in the investigation of medical images and is predominantly implemented in various medical applications. In the case of investigating MRI brain images, the image segmentation is mainly employed to measure and visualize the anatomic structure of the brain that under-went modifications to delineate the regions. At present, distinct segmentation approaches with various degrees of accurateness and complexities are available. But, it needs tuning of various parameters to obtain optimal results. The tuning of parameters can be considered as an optimization issue using a similarity function in solution space. This paper presents a new Parametric Segmentation Tuning of Canny Edge Detection (PST-CED) model. It is based on the comparison of consecutive segmentation outcomes and the selection of one that yields maximum similarity. Besides, this paper employs an effective pre-processing technique at an earlier stage, i.e., before segmentation to improve the image quality. Here, a hybrid contrast stretching approach was employed depending upon the Top-hat filter and Gaussian function. The PST-CED technique was tested with benchmark MRI images and a detailed comparative analysis was conducted with state-of-the-art methods interms of Peak Signal to Noise Ratio (PSNR), detection accuracy, execution time and Mean Square Error (MSE).