Automatic Detection of Tumor in Wireless Capsule Endoscopy Images Using Energy Based Textural Features and SVM Based RFE Approach

This paper deals with processing of wireless capsule endoscopy (WCE) images from gastrointestinal tract, by extracting textural features and developing a suitable classifier to recognize as a normal or abnormal /tumor image. Images obtained from WCE are prone to noise. To reduce the noise, filtration technique is used. The quality of the filtered image is degraded, so to enhance the quality of the image, discrete wavelet transform (DWT) is used. The textural features (average, energy) are obtained from DWT for three color spaces (RGB, HSI, Lab). Feature selection is based on support vector machinerecursive feature elimination approach. KeywordsDiscrete Wavelet Transform (DWT), Feature selection, Support Vector Machine (SVM), Texture, Wireless Capsule Endoscopy (WCE).