Indistinct Frame Detection in Colonoscopy Videos

Endoscopic Video Manifolds for Targeted Optical Biopsy

“…Once the rotation-invariant energy histogram of each frame is computed, the similarity between two histograms is measured based on the cosine similarity measure (15) where is the dot product and denotes the norm of the dimensional histogram vector. Due to the dot product formulation, the cosine similarity relies on the angle between the two histogram vectors in the -dimensional space.…”

“…Presence of these uninformative frames leads to poor classification performance. In the literature, detection of uninformative frames has been studied for several endoscopic procedures such as capsule endoscopy [12]- [14], colonoscopy [15], [16]. Main focus of these studies is on defining specific features such as color or texture in order to detect uninformative frames within an endoscopic video.…”

Content-based processing and analysis of endoscopic images and videos: A survey

“…The most important criterion is blurriness. According to [9], about 25 % of the frames of a typical colonoscopy video are blurry. Oh et al [169] propose to use edge detection and compute the ratio of isolated pixels to connected pixels in the edge image to determine the blurriness.…”

“…Seven texture features are extracted from the gray-level co-occurrence matrix (GLCM) of the resulting frequency spectrum image and used for k-means clustering to differentiate between blurry and clear images. A similar approach by [9] uses the 2D discrete wavelet transform with a Haar wavelet Kernel to obtain a set of approximation and detail coefficients. The L 2 norm of the detail coefficients of the wavelet decomposition is used as feature vector for a Bayesian classifier.…”

Towards Intelligent Systems for Colonoscopy

“…The work of Arnold et al (2009) addresses the identification of clinically uninteresting frames by analyzing the energy of the detail coefficients of the wavelet decomposition of a given image, which is used as the input to the classification system. In this case non-informative frames are those which do not carry any useful clinical information, such as those that occur when the camera is covered with liquids or when it is very close (even touching) the mucosa.…”