An automatic and intelligent computer aided bleeding detection technique is presented to recognize the bleeding regions and other pathological features in large amounts of images generated from a Wireless Capsule Endoscope(WCE).Color features of the bleeding region in WCE images is extracted, and then the Probabilistic Neural Network (PNN) is improved by using differential evolution (DE) algorithm to offer the different smoothing parameters for each transfer function of neurons.Based on the improved PNN, the intelligent recognizing method is proposed and implemented through programming.The experimental results show that the bleeding regions in WCE images can be recognized correctly and marked clearly, and the sensitivity and the specificity of the method are measured as 94% and 87%, respectively.The intelligent bleeding detection method reduces the time-consuming for the WCE video detection and can help the clinician examine the gastrointestinal disease.