HEVC Thumbnail Extraction Device and Method

Low-Complexity HEVC Thumbnail Extraction Device Based on Intra-Prediction Mode Weight Tables

This technology is a device and method for generating thumbnail pixels by performing an inverse discrete cosine transform on an HEVC bitstream, partially reconstructing only the residual values of specific 4×4 boundary pixels, and calculating and adding weighted average prediction values and weighted average residual values using weight tables configured according to intra-prediction modes.

Thumbnail extraction algorithms for MPEG-2 or H.264/AVC rely on DC value extraction or hundreds of 2D lookup tables, making them unsuitable for HEVC, which utilizes planar modes, various intra-directions, and large block sizes. Adapting them to HEVC would require storing 1,680 lookup tables and would result in significant image quality degradation in large blocks.

This technology minimizes computational complexity and effectively eliminates aliasing by generating thumbnails through partial transformation and mode-specific weight calculations without full decoding. It can be applied to video streaming services, video search/preview features, and digital broadcast content management systems, enabling fast and clear generation of thumbnail lists for high-capacity HEVC videos.

‍

‍

Key Features:

• A partial decoding unit that performs an N×N inverse discrete cosine transform on an HEVC-compressed video bitstream and reconstructs the residual values of specific 4×4 boundary pixels
• A block generation unit that creates a prediction block in the intra-mode from 4×4 pixels and generates a residual block by subtracting the prediction block from the current block
• A down-sampling unit that calculates weighted average prediction values and weighted average residual values using a predefined weight table and adds them to generate thumbnail pixels
• A weight table where weights corresponding to each average residual value are set differently according to the intra-prediction mode

‍