Abstract
With the commercialization of data broadcasting through communication and broadcasting convergence, interactive multimedia services providing additional information about on-screen objects have become possible. A critical factor in these services is user-centric interoperability, which allows for immediate information retrieval upon user request. This paper proposes an efficient object storage method to enhance this interoperability. The proposed method identifies the "Object’s Own Area" (OOA) within a mask video, partitions it into equal blocks, and clusters these areas into bit-sum formats based on a minimum object density threshold. Experimental results demonstrate that this approach significantly reduces the storage space required for object position information compared to traditional methods, improving overall system efficiency.
Introduction
In the era of interactive broadcasting, viewers expect to engage directly with content—for instance, clicking on a product in a commercial or an ingredient in a cooking show to access more information. Enabling this functionality requires the system to track the precise coordinates of interactive objects across every video frame. Traditionally, this is achieved using mask videos, where the object of interest is isolated in white against a black background. However, high-definition mask videos impose massive storage and bandwidth burdens, often exceeding the processing capabilities of standard Set-Top Boxes (STBs).
The BitSum Clustering for Mask (BSCM) Algorithm
To address these storage challenges, we propose the BitSum Clustering for Mask (BSCM) algorithm. This method shifts from storing raw pixel-based masks to a structured, compressed data format that is both space-efficient and computationally light for real-time interaction.
The BSCM process consists of five primary stages:
- Object’s Own Area (OOA) Extraction: The algorithm first identifies the smallest rectangular bounding box that encompasses the entire object, ignoring the rest of the frame.
- Equal Partitioning: The OOA is divided into a grid of blocks, each of size $D \times D$. This grid-based approach simplifies the spatial representation of the object.
- Density Filtering (Min_Rate): For each block, the algorithm calculates the ratio of object pixels to total pixels. If this density is below a predefined threshold ($Min\_Rate$), the block is discarded. This removes insignificant edge details to save space.
- BitSum Clustering: The indices of the remaining "active" blocks are stored as bitsets (BitSums). This allows the STB to perform extremely fast bitwise operations to determine if a user's click coordinates fall within an interactive region.
- Temporal Sampling (Frame Interval): To further reduce data volume, the algorithm employs a Frame Interval ($FI$) parameter, storing object data only every $N$ frames for objects with predictable motion.
Experimental Results and Discussion
The proposed method was tested using HD (1280x720) mask videos. The experiments focused on the trade-off between storage efficiency and object representation accuracy.
- Spatial Compression: By utilizing a $20 \times 20$ block grid instead of raw pixel data, the spatial information was reduced to approximately 1/40th of its original size.
- Temporal Compression: Setting the Frame Interval ($FI$) to 10 effectively reduced the data volume by 90% without significantly impacting the user experience for moderately moving objects.
- Parameter Sensitivity: Increasing the $Min\_Rate$ threshold led to a decrease in the number of stored clusters, demonstrating that broadcasters can tune this parameter based on their specific storage and precision requirements.
Conclusion
The BSCM algorithm provides a robust framework for managing interactive object data in multimedia broadcasting. By combining spatial gridding with bit-sum clustering and temporal sampling, the method achieves high compression ratios while remaining hardware-friendly for low-power Set-Top Boxes. Future research will explore handling complex scenes with multiple overlapping objects and dynamically adjusting parameters based on object velocity to further optimize the balance between precision and performance.
Summary updated