In the field of image night vision, the registration accuracy of multi-model images might have a significant effect on the final fusion image quality. Due to the lack of the image registration objective evaluation method which conforms to subjective feeling, we researched on the objective evaluation index of multi-sensor image registration based on boundary definition, target saliency and color consistency. Through the introduction of weber's law, we built the human eye perception model. Combined with the local band contrast model, we proposed the boundary definition index of gray fusion image. At the same time, we used the frequency-tuned salient region detection model to calculate the target saliency index of the color fusion image. We calculated the color difference between the fused image and reference color image, and then proposed a color consistency index of image registration. At last, we proposed a new registration evaluation method for multi-model images which has superior performance. The experimental results showed that the new index has higher consistency with the subjective feeling and this index could effectively assess the multi-model images registration accuracy problem in the night vision field.
A method for moving target detection and segmentation using Markov random field (MRF)-based evaluation metric in infrared videos has been proposed. Starting with the most useful seeds of a moving object, which are extracted based on the “holes” effect of temporal difference; the proposed method employs a region growing method using local gray information and a spatial and temporal MRF model-based evaluation metric without ground truth for moving target segmentation in infrared videos. The segmented mask of a moving target is grown from the most useful seeds using the region growing method with thresholds. The proposed evaluation metric is utilized to determine the best growing threshold, where the performance of moving target segmentation is measured by that of segmented mask’s boundary. Thus, an MRF modeling for each boundary point of the segmented mask in spatial and temporal directions was considered by us. This problem is formulated using maximum a posteriori (MAP) estimation principle. At last, the global optimum of MRF-MAP framework is achieved using simulated annealing algorithm. The best segmented mask of a moving target is grown from the most useful seeds with the best growing threshold. Experimental results are reported to demonstrate the accuracy and robustness of our algorithm.
Proc. SPIE. 9142, Selected Papers from Conferences of the Photoelectronic Technology Committee of the Chinese Society of Astronautics: Optical Imaging, Remote Sensing, and Laser-Matter Interaction 2013
To find out the best infrared and visible fusion system of fusion algorithm which has excellent target detection characteristics in different environment, we proposed a new fusion algorithm selective rule. We also defined new concepts: fusion algorithm coefficient and the equivalent transmissivity of system. Using local-target contrast, local-target articulation to calculate fusion algorithm coefficient, we can estimate the target detection performance of fusion system when it working in different air humidity environment. Also, we make use of infrared and visible fusion system designed by ourselves to verify this method. Besides fusion algorithm coefficient, we also use subjective evaluation to evaluate the target detection performance of fusion algorithm. At last, the best algorithm or the method which is most consistent with human visual in different conditions were found. Through this work, we can provide the basis for the algorithm of choice in the fusion system.
A new method for unsupervised segmentation of moving objects in infrared videos is presented. This method consists of two steps: difference image quantization and spatial segmentation. In the first step, the changed pixels in the difference image are quantized to several classes by using Bayes decision. It can be used to cluster the changed pixels belonging to the same moving object together. The pixels of the difference image are replaced by their corresponding class labels, thus forming a class-map of the difference image. In the second step, each class in the class-map is considered as a subset of the possible seeds of moving objects. A self-adaptive region growing method is then used to image segmentation on the basis of these different subsets. One of the focuses of this work is on spatial segmentation, where a criterion is proposed for evaluation of moving object segmentation without ground truth in infrared videos. This criterion is used to evaluate the performance of the segmentation masks grown from different subsets of the possible seeds. The best segmented image is determined to be the final segmentation result. Experiments show the advantage and robustness of the proposed algorithm on real infrared videos.