Algorithm and VLSI Architecture of Edge-Directed Image Upscaling for 4K Display System (submitted to TCSVT)

 
This paper proposes an explicit edge-directed adaptive image interpolation method that leverages more sophisticate edge detection and orientation estimation algorithms to avoid the mis-interpolation, thereby providing the similar or even better image quality with respect to implicit interpolation methods. Targeting for the real-time 4K video display system, the proposed edge-directed image upscaling algorithm is further implemented with an efficient VLSI architecture. Experimental results demonstrate that the proposed interpolation algorithm outperforms the previous explicit and implicit edge-directed methods in both objective and subjective test. The presented VLSI implementation further demonstrates that the maximum output video sequence of the proposed interpolation method can reach up to 4Kx2K@60Hz with reasonable hardware cost.

The set of images (both classic images and from the morgueFile archive) used in our experiments is available following this link.

Results of 4x enlargements for bicubic interpolation are available following this link.

Results of 4x enlargements for the proposed explicit edge-directed interpolation are available following this link.