Undecimated Complex Wavelet Transform based bleeding detection for endoscopic images

Abstract

Wireless Capsule Endoscopy is a methodology to detect abnormalities in the Gastrointestinal (GI) tract, mostly the internal regions of small intestine. It is a non-invasive process. In this work, Undecimated Double Density Dual Tree Discrete Wavelet Transform (UDDT-DWT) is considered in detecting bleeding WCE images. Four statistical parameters such as contrast, entropy, cluster shade and cluster prominence are calculated from Gray Level Co-occurrence Matrix (GLCM) of each sub images obtained after applying UDDDT-DWT. These features are used for the classification of WCE images. For the detection of blood in images, endoscopic image is converted to HSV colour space and several classifiers are considered. Experiments show that the proposed method provides a high accuracy rate of 99.5%, sensitivity of 99% and specificity of 100% for Random Forest and Random Tree classifier when compared with the existing methods.