In high-resolution image segmentation tasks, applying graph-based algorithms can be computationally intensive and prone to performance bottlenecks. So, we propose a saliency-driven, tile-aware segmentation framework for high-resolution traffic surveillance imagery. Traditional full-frame segmentation methods often struggle with computational inefficiency and loss of detail in object-dense regions. To address this, we introduce a pipeline that first detects salient regions using a fine-grained saliency map, followed by strategic tiling that prioritizes these regions. Each tile is then segmented independently in a parallel multicore setup, significantly improving processing speed. Finally, the segmented tiles are merged into a coherent segmentation map with careful overlap handling to ensure continuity across boundaries. Experimental results demonstrate that our method preserves object boundaries more accurately in dense scenes while maintaining computational efficiency. This approach is particularly well-suited for real-time traffic monitoring and other surveillance applications where scalability and precision are critical.