Search

This paper presents a design of new video coding scheme targeting substantial compression performance with reasonable complexity for next-generation high-resolution video sources. While it takes a conventional block-based MC+DCT hybrid coding approach that is suitable for hardware implementation of a high-resolution video codec, the proposed scheme achieved significant improvement of coding efficiency by introducing various technical optimizations in each functional block, especially allowing the use of larger blocks for motion compensation and transform than conventional standards. According to our experimental analysis, the proposed scheme achieves approximately 26% bit-rate savings on average compared to the state-of-the-art standard AVC/H.264 high profile. We also study the tradeoff between complexity and performance of the codec towards its standardization and implementation.

In this paper, we propose a novel text line extraction method using domain-based optimization for complex documents such as engineering drawings. In the complex documents, variations of text line due to differences in size, interval, and direction, as well as the complex layout (text lines and drawing objects are mixed) make the problem of automatic text line extraction extremely challenging. The proposed method divides a document into some domains based on line objects or spaces. Then, this method generates a potential map in each domain. Finally, text lines are extracted by an active contour model based on the potential map. In this process, the potential map is generated by a potential function that has some parameters concerning interval, character size and drawing objects. By optimizing these parameters, this method extracts text lines easily in complex documents. Our experimental result shows better performance than a conventional method for engineering drawings.

This paper proposes a new method for improving direct prediction scheme that has been employed in conventional video coding standards such as AVC/H.264. We extend direct prediction concept to achieve better adaptation to local statistics of video source with the assumption of the use of larger motion blocks than conventional macroblock size. Firstly, our direct prediction method introduces block adaptive spatio-temporal estimation of direct motion vector, in order to compensate loss of accuracy of motion vector estimation for large motion blocks. This estimation method is performed without explicit signaling by employing decode-side collaborative decision. Then, adaptive selection of two reference pictures is performed to improve direct prediction efficiency where reliability of estimated direct motion vector is poor. Experimental results show the proposed method provides up to 3.3% bitrate saving and 1.5% in average in low-bitrate coding.