Feedback-Guided Block-Level Encoding Prioritization Using Bandit Learning for Image Transmission

Block-based image transmission is an effective approach for efficient bandwidth utilization, as it allows the sender to transmit only those regions of an image that carry the most relevant visual information. In this framework, an image is partitioned into multiple non-overlapping blocks, each containing different spatial details and levels of perceptual importance. However, the perceived importance of image regions varies among users, depending on their viewing preferences or application requirements. Such diversity creates a need for an adaptive mechanism that can dynamically determine which blocks deserve higher transmission priority. To address this issue, we introduce a feedback-guided prioritization algorithm based on the multi-armed bandit (MAB) model. The proposed algorithm updates the priority of each block according to user feedback by learning the statistical relationship between transmitted content and user satisfaction. Through this continuous interaction, the system gradually identifies the regions that contribute most to user-perceived quality and allocates more transmission resources. Comprehensive simulations verify that the proposed framework can effectively capture user preferences and enhance the overall visual experience.