A Hybrid Deep Composition Optimization Approach for Real-Time Portrait Photography Assistance

Portrait photography plays a big part in digital communication today. The challenge for the amateur user is to get a nice looking composition because of the lack of knowledge of the principles of photography. Here we propose a Hybrid Deep Composition Optimization approach for real-time portrait photography assistance that combines deep learning with rulebased composition analysis. The system we propose takes live camera frames and applies a lightweight face-detection model (BlazeFace) to infer the position and orientation of the subject. Specialized computer vision modules estimate several composition constituents, such as rule of thirds, spatial balance, lighting consistency, and visual saliency.To quantify overall composition quality, we propose an Adaptive Composition Similarity Index (ACSI) that combines these factors into a unified scoring framework for real-time feedback. Furthermore, the system uses MobileNetV2 for deep feature extraction for reference-based composition matching. Experimental results show that the proposed hybrid approach achieves an accuracy of 93% at the same time maintaining the real-time responsiveness, outperforming the conventional CNNbased methods.The proposed system provides actionable guidance including subject repositioning, camera alignment and lighting adjustments so that users can take visually balanced portraits with no prior expertise. This approach provides a scalable solution for intelligent photography assistance on smartphones, smart cameras, and edge-based imaging systems.Unlike conventional post-processing approaches, the proposed system provides composition guidance during image acquisition in real-time.