基于人眼视觉的任意形状区域丢失图像信息重建

任意形状区域丢失图像信息的重建对于图像中遮挡区域的去除、受损图像的复原和红外图像中的目标检测具有十分重要的意义。目前普遍采用的插值重建等方法存在重建后图像的主观视觉质量明显下降的不足。因此提出了基于人眼视觉特性的任意形状区域丢失图像信息的重建模型。利用重建区域外接矩形中余留的已知信息 ,并根据人眼的低通视觉特性 ,以加权最小二乘的方式在DCT域中重建相应的DCT系数 ,然后通过IDCT得到重建后的图像。实验证明 ,通过该方法所得到的任意形状区域的重建图像 ,在主观和客观质量方面都有显著的提高。

Reconstruction for the Lost Image Information of the Arbitrary-Shape Area Based on Hmnan Visual Characteristics

The reconstruction for the lost image information of the arbitrary-shape area is very important for the disocclusion and restoration of the image and the object detection in the infrared image. In general, the widely used interpolation method has some limitations such as low subjective quality of the reconstructed image. The reconstruction model for the lost image information of the arbitrary-shape area based on the human visual characteristics is put forward. According to the low-pass characteristics of the human visual system, the remained known information in the circumscribed rectangle of the arbitrary-shape area is used to reconstruct the discrete cosine transform (DCT) coefficients in the DCT domain by the weighted least square method. After the inverse discrete cosine transform (IDCT), the reconstructed image can be obtained. Many experiments demonstrate that the subjective quality and objective quality of the reconstructed images are greatly improved by this method.