Adaptive DoFP polarization image demosaicking based on local gradient and channel correlation

With the advancement of nanotechnology, division-of-focal-plane (DoFP) polarization imaging systems with real-time imaging capabilities have emerged as a significant research area, advancing the application potential of miniaturized polarization imaging systems. Nonetheless, the superpixel structures within these systems can introduce instantaneous field-of-view (IFoV) errors, which affect the quality of polarization image reconstruction and the accuracy of polarization information calculations. Most existing polarization image demosaicking methods based on channel correlation assume uniform scene distribution, utilizing fixed weights or thresholds. This reliance produces poor robustness across various scenes, making these methods less suitable for practical applications. To address these limitations, this paper proposes an adaptive DoFP polarization image demosaicking method based on local gradient and channel correlation (ALGPCC). Specifically, the method first employs local gradient optimization on the traditional bilinear interpolation method to produce a high-quality initial demosaicked image. Next, it combines normalized cross-correlation with guided filtering to create adaptive polarization channel correlation weights, allowing for dynamic adjustment based on the polarization characteristics of various scenes. Finally, these adaptive weights are applied to a polarization channel difference model, further improving the demosaicking results and effectively reducing IFoV errors. Experimental results with synthetic and real DoFP polarization images demonstrate that the proposed method significantly surpasses existing demosaicking methods in objective metrics and visual quality, showing superior performance across various scenes and offering notable advantages in processing speed.