Wireless capsule endoscopy (WCE) has revolutionised the diagnosis and treatment of gastrointestinal tract, especially the small intestine where traditional endoscopies cannot reach. However, this new technology leads to the inspection of a large number of images, which is a time-consuming process and also too hard by naked eyes for doctors. In this paper, we propose a new computerised method for bleeding detection in WCE images. We use the second component of CIE Lab colour space together with appropriate segmentation and enhancement techniques, involving an adaptive anisotropic diffusion (alike Perona–Malik diffusion). As a result of this procedure, it is possible to devise four functions to discriminate between bleeding and normal regions in WCE images. These four bleeding detectors rely on the eigenvalues of the Hessian and on the Laplacian of the modified enhanced image. Multiscale image analysis approach is also involved in the definition of these detectors for handling the maximum and minimum sizes at which the bleeding regions are expected to be found. Experimental results on several medical data-sets show that the new algorithm achieves a very good rate of success and promising performance for bleeding detection.