Hybrid Approach for Shot Boundary Detection Based on Machine Learning

High level video generalization, video segmentation, video indexing, video summarization and video recovery required an elementary step of recognition of shot boundaries. Therefore detection of shot boundaries is prerequisite for revealing content of complicated video structure. There are several applications where shot boundary detection automation can be used such as Geographical information system (GIS), restoration of video, multimedia news, digital libraries, Tele-learning, interactive display. There are some problems that still need to be addressed by the researchers to resolve. The disturbance caused by the illumination change, detection of gradual and abrupt transitions is the main confronts in the detection of shot breaks. In this book we propose three methods for shot boundary detection which are Dual Tree Discrete Wavelet Transform (DTDWT), Artificial Neural Network (ANN) and Convolution Neural Network (CNN). We have assessed algorithms using performance evaluation metric Precision, Recall and F1 measure.

Autorentext
Dr. Neelam Labhade-Kumar has obtained her Ph.D in Electronics and Communication Engineering. Her area of research is Video Shot Boundary Detection, Image Enhancement and Retrieval. She has published papers in several reputed journals like Springer, Elsevier, Scopus and UGC care etc. She has total 14 years of teaching experience.