Holoscopic imaging, also referred to as integral imaging, is an advantageous solution for glassless 3D which promises in the future to change the market for 3D video systems. In order to efficiently transmit this type of 3D video content over current and emerging networks, this paper proposes an improved spatial and temporal prediction scheme which attempts to exploit the particular spatial self-similarity of this type of content, as well as the high temporal correlation between successive frames, in order to achieve a better prediction and, consequently, improve the coding performance. Initially, this paper provides a brief presentation of the general concepts behind holoscopic imaging with special attention to the spatial correlations, which are inherent to this type of content, due to the micro-lens array used both for acquisition and display. This acquisition process generates a planar intensity distribution behind the micro-lens array that contains an array of micro-images in which a different perspective of the scene is shown. These micro-images exhibit a high correlation between them, which can be seen as a form of self-similarity within the holoscopic image. Experimental results based on a modified H.264/AVC video codec that can handle 3D holoscopic self-similarity estimation and compensation are presented and clearly show advantages over H.264/AVC, in terms of coding efficiency of using the proposed approach for full parallax 3D holoscopic content.