Hybrid Quantum-Classical Convolutional Squeeze Excitation Neural Network based HEVC Video Quality Enhancement

Due to the popularity of video streaming on the internet, there is a high utilization of HEVC because of the high compression ratio it delivers in video quality. However, the compression process causes distorting elements that have a negative impact on the quality of the videos, which presents much complication in preserving the content integrity as it was decoded from the videos. To overcome these issues, this manuscript proposes a Hybrid Quantum-Classical Convolutional Squeeze Excitation Neural Network for HEVC Video Quality Enhancement through a two-phase approach: There are two areas namely, (i) Quantization Bit Stream and (ii) post-processing. In the first phase, for example, a Squeeze Excitation Multi-Head Attention-Based Encoder and Decoder radically effects a video frame compression by utilising possibilities of prediction, transformation, quantization, and entropy coding, though it distorts at the same time. The decoder then unpacks the frames thus created to the format that the enhancement module uses as input. The second phase makes use of a Hybrid Quantum-Classical Convolutional Neural Network (HQ2C2N). This approach involves a Feature Extraction Layer that employs quantum circuits to optimize the feature extraction step, a Feature Enhancement Layer that applies classical convolutional networks to eliminate artifacts that have been introduced to the video and a Post-Processing Layer that allows for tuning of end frames to the desired quality standards typical of plyback or further processing. The experiments of the proposed HQ2C2N method are conducted using MATLAB programming environment. The HQ2C2N significantly improves HEVC video quality by effectively reducing compression artifacts, resulting in a higher PSNR of 75 dB and a lower MSE of 0.1%, outperforming previous methods in visual fidelity and performance metrics.