Cross-Modality Compression Efficiency and Reconstruction Fidelity in Mobile Cloud Computing: A Unified Runtime Analysis of the EADC Framework

Abstract Mobile cloud computing backends must compress heterogeneous data under strict latency and bandwidth constraints, yet compression studies usually evaluate one data type and report runtime only as an aggregate. This paper presents a unified runtime and reconstruction-fidelity evaluation of the Enhanced Adaptive Discriminant Compression framework across its image and text pathways. An additive runtime model attributes processing time to kernel and discriminant computation, dictionary-based redundancy elimination, and adaptive parameter updates. Experiments on eight image categories from the Stanford Light Field Raw Image Dataset and eight text collections from the Pizza & Chili Repetitive Corpus show reconstruction accuracy between 99.90% and 99.98%. Image-pathway runtime differences against dictionary and context-model baselines are statistically significant (p < .001). Mean total runtime is 3.71 s per image and 8.01 s per text collection, corresponding to throughputs of 11.2 MB/s and 58.4 MB/s, which supports near-real-time deployment on cloud backends.