With the explosion of multimedia data (videos, audios, images, and Web pages), people have no time to look at this huge data, and human attention has become a precious resource. So, a way must be found to automatically analyze, classify, summarize, discover and characterize trends in it, and to automatically flag anomalies. Many researchers have felt the need for data mining methods to deal with this amount of data. The Multimedia Mining work includes various kinds of data like video, audio, image, etc. This work includes a set of fields, which process huge information, like clustering, recognition, classification, etc. Video Mining is the most important kind of mining because the video became a popular easy way to deliver a message or an idea like the video chat, movies, advertisements and also a way for protection like insurveillance videos and a lot of other uses that the world is full of. In this research, a Video Retrieval System (VRS) was implemented.This system has two phases: enrollment phase and query-matching phase.The enrollment phase constructs the videos' database. This database contains different classes that clustered according to their textural (Run Length) or shape (Moment Invariant) features with different color models (HSL or YCbCr).The clustering process is accomplished using K-means clustering algorithm. While, query-matching phase used for retrieving videos which are similar to an input one. Before extracting features, the input video is cut into several shots depending on the difference between successive frames histograms, then calculating the features to a chosen shot to retrieve the similar shots from the videos' database.The performance of the system was computed using recall and precision measures. In VRS, an encouraged good result was obtained.