In the present era, the social network is used as an important medium for sharing thoughts and opinions of an individual. The main reason behind this is, it provides a fast-spreading of information among the public easily, requiring a very low cost of access. This leads to having online social media as one of the stepping stones to encourage false content and influencing public opinion and its decision. Rumour is one of the prominent forms of misleading information on social media and should be detected as early as possible for avoiding their significant effects. Due to these reasons, the researchers have put their keen interest in developing an effective rumour detection framework in the last years. In this paper, we mainly focused on six main aspects. Firstly, we discuss rumours from a definition perspective that have been considered in the state-of-the-art and describe the generalized model of rumour detection. Secondly, we discuss how to get access to data from different social media platforms, and presents various state-of-the-art methods to gather these data, as well as publicly available datasets. Third, we describe a different set of features that have been considered in rumour detection approaches. Fourth, we provide deep insight into the various methods used to employ rumour detection and its veracity assessment on multimedia data (Text and Images) with some practical implications. Whereas in the fifth aspect, the constraints of the study have been discussed. Finally, we concluded with useful findings and suggested future directions.