Attention-driven low-cost image contrast enhancement for low-power OLED display

Organic light-emitting diode (OLED) displays demand a high computation cost to process the high-resolution content without optimizing the power usage. This paper presents a power-aware visual contrast enhancement (PAVE) method that efficiently processes high-resolution images, reduces computation costs, and optimizes the display's power. The proposed method utilizes original and low-resolution images to extract multilevel features. A power-attention mechanism, fast-guided filter (GF), and local tone mapping are incorporated to preserve on-demand power and reduce computational cost. Based on thorough experiments, our method yields an optimized image with an SSIM of 0.98% and a CPU time of 0.05 seconds at 2 k resolution.