光场成像作为一种应用广泛的计算成像方法,其研究仍局限于电磁波谱中的可见光波段。因此,提出基于相机阵列的红外光场成像方法。首先利用单台被动红外相机移动到同一平面的不同位置分别对同一静态目标场景成像,获取原始红外图像序列,经平移视差校正后生成红外光场数据库;然后基于光场渲染理论,获得重采样图像。与只记录二维位置信息的传统红外成像比较,红外光场成像能够记录包含位置和方向的四维信息。实验结果表明,该法能够融合多幅低信噪比的图像获得一幅信噪比提高的重采样图像,通过孔径叠加平均可实现穿透遮挡物成像,通过选取不同深度值进行重采样可实现红外场景不同深度的数字重聚焦,使红外成像在**和民用领域更有应用价值。

As a widely used computational imaging method, the light field imaging's research is still limited to the visible band in the electromagnetic spectrum. Based on the camera array, a method of infrared light field imaging is proposed. A passive infrared camera is used to obtain the original infrared image sequences of the same static target at different positions of the same plane. The infrared light field database is built after translational parallax correction. Then resampling images are obtained based on the light field rendering theory. Compared with the traditional infrared imaging which only records the two-dimensional position information, the infrared field imaging can record four-dimensional information of the position and direction. Experimental results show that this method can obtain a resampling image with higher signal-to-noise ratio by blending multiple low signal-to-noise ratio images, realizing the imaging of the penetration barrier by the aperture superposition average and achieving the digital refocusing with different depths of infrared scenes by resampling different depth values selected. The proposed method can make the infrared imaging more valuable in military and civilian fields.