Brain cancer is a major health problem that affects the lives of many people. Clinicians depend on their medical knowledge to analyse laboratory test results and clinical information extracted manually from medical images, to identify the essential characteristics of tumours such as size, shape, and location. The accurate diagnosis of brain tumours is very essential for deciding the most appropriate treatment protocols and procedures. Apart from being tedious, slow and time-consuming manual segmentation of brain image scans is also prone to human error. In order to enhance the accuracy and reliability of the diagnosis of brain tumours, rigorous image processing and segmentation is required. In this study the authors present an automated image segmentation method for detecting and localising multiple brain tumours of different sizes and intensities by using a technique of object counting coupled with an evaluation process. The object counting technique were carried out using multiple binary images with different threshold values, and the evaluation process was based on assessment criteria of specific tumour parameters. Different magnetic resonance images with multiple tumours were used for testing the proposed method. The results indicated that the proposed method is effective in detecting and localising multiple true tumours in the brain.