| The key component of today's infrared imaging systems is the uncooled infrared focal plane array.However,due to the limitation of manufacturing process and material,the image obtained by the focal plane array has serious non-uniformity,especially the stripe non-uniformity seriously affects the quality of the image.In addition,the collected infrared image has the disadvantages of low contrast,unclear image details,etc.,moreover due to the high dynamic range of the original image,it is necessary to compress the bit depth to 8-bit,which adapts display device.The compression process will lose some details,which limits the subsequent applications such as detection and recognition of the target.Therefore,aiming at the above problems,this paper gives a solution at the algorithm level.The main work is as follows:Firstly,this paper introduces the midway histogram equalization algorithm,which directly uses the neighborhood information in the vertical stripe direction to handle the stripe non-uniformity adaptively.The method based on the midway histogram can accommodate more scenes and get good results in most pictures.However,the processing speed is slower,cannot remove stripes completely in flat areas,and the other noise cannot be processed.So a method of finding the stripe pattern and then using the original image to reduce the stripe to obtain the correction result is mainly studied.That is,the stripe non-uniformity correction algorithm based on one-dimensional guided image filtering.The edge blurring is caused by the algorithm mistakenly streaking long edges that are consistent with the stripe direction.In this paper,the edge detection operator is used to obtain the vertical edge index graph,and the graph is used to guide the smoothing coefficient of row guided image filtering with a smaller one at the vertical edge point.And if there is a longer edge,the larger filter window radius is selected in the column guided image filtering.In addition,according to the scene adaptive adjusts the guide filter coefficients to solve the problem of manual parameters.The improved method can better deal with the stripe while reducing the edge blur,while being able to have a certain removal effect on the other kind of noise in the image,but the brightness and contrast of the image have a partial loss.Secondly,this paper analyzes the digital detail enhancement technology(DDE)based on edge-preserving filter layering.Firstly,the bilateral filtered digital detail enhancement algorithm is analyzed.However,in the bilateral filter layering,some significant edge information is missing to the detail layer,which make the gradient inversion effect appear.At the same time,a large number of distortion artifacts are caused during the enhancement of the detail layer.In this paper,a gradient penalty term is added to the edge-preserving filter energy function to distinguish the strong and weak edges,and a new gradient-perception edge-preserving filter is obtained.The strong edge and the overall brightness of the image can be preserved in the DDE-separated base layer.Only the weak edges and details are enhanced as the detail layer,effectively preventing distortion such as gradient inversion and halo effect.Moreover,according to the theory of human vision,the gradient and brightness weights are designed to enhance the details preventing over-enhancement during the enhancement of the detail level.Finally,in order to display the details of the image as completely as possible on an 8-bit display,this paper uses an improved histogram projection to compress the dynamic range.The subjective perception and objective evaluation indicators of the image show that the proposed algorithm can deal with the stripe non-uniformity of well,ensuring that there is no blurry edge or other distortions.A new DDE method proposed by this paper can enhance the detail while compressing the dynamic range of the image without introducing distortion.It lays the foundation for subsequent infrared image applications. |
PDF Full Text Request