Bingjian Wang

Visible and infrared image registration based on visual salient features

Abstract. In order to improve the precision of visible and infrared (VIS/IR) image registration, an image registration method based on visual salient (VS) features is presented. First, a VS feature detector based on the modified visual attention model is presented to extract VS points. Because the iterative, within-feature competition method used in visual attention models is time consuming, an alternative fast visual salient (FVS) feature detector is proposed to make VS features more efficient. Then, a descriptor-rearranging (DR) strategy is adopted to describe feature points. This strategy combines information of both IR image and its negative image to overcome the contrast reverse problem between VIS and IR images, making it easier to find the corresponding points on VIS/IR images. Experiments show that both VS and FVS detectors have higher repeatability scores than scale invariant feature transform in the cases of blurring, brightness change, JPEG compression, noise, and viewpoint, except big scale change. The combination of VS detector and DR registration strategy can achieve precise image registration, but it is time-consuming. The combination of FVS detector and DR registration strategy can also reach a good registration of VIS/IR images but in a shorter time.

A new small and dim targets detection and recognition algorithm based on infrared dual bands imaging system

In this paper, infrared radiation characteristics in different bands of flight targets, background and decoys are analyzed. According to the analytical results, a new small and dim targets detection and recognition algorithm based on infrared dual bands is present. Infrared dual bands imaging system includes infrared mid-wavelength imaging system and infrared long-wavelength imaging system. Images from these two imaging system are registered in time and space firstly. After background suppression, then the small and dim targets are detected from MW infrared images and LW infrared images respectively. At last, by using spatial correlation and spectral correlation of targets, interferes and decoys are eliminated from target candidates. The genuine targets are separated and recognized. The algorithm is applied to simulation dual bands images and real infrared dual band images. The experimental results show that the algorithm present in this paper is effective.

A Novel Non-Local Means Based Super-Resolution Algorithm with Back-Projection

A novel multi-frame super-resolution (SR) algorithm without explicit motion estimation is proposed. This method is intended to promote the precision of SR reconstruction by embedding the back-projection technique into the non-local means (NLM) based fusion and reconstruction process. Results on test movie show that the proposed method is very successful in providing super-resolution on general sequences and is able to remove the bur and edge artifacts effectively.

Registration of infrared and visible images based on the correlation of the edges

Registration of infrared and visible images is very difficult because of their different imaging principles. Considering the correlation of edges in these two kinds of images, an improved registration algorithm is proposed in this paper. Firstly, the wavelet transform modulus maximum algorithm is used to detect edges in images. Then the Speeded Up Robust Features (SURF) algorithm is used for feature points detection on the edges. Finally, feature points are matched by rough matching and accurate matching, and the least squares method is employed to find optimal solution to affine transform equations. Experimental results show the high accuracy and stability of this registration method.

Shearlet transform based anomaly detection for hyperspectral image

Hyperspectral image (HI) contains data in hundreds of narrow contiguous spectral bands, thus it provides a powerful means to distinguish different materials on the basis of their unique spectral signatures. Anomaly detection (AD) is one key part of its application. The shearlet transform (ST) is a new two-dimensional extension of the wavelet transform using multiscale and directional filter banks, which can effectively captures smooth contours that are the dominant feature in natural image. In this paper, ST is used in AD for the HI. Firstly, the raw HI data is decomposed into several directional subband at multiple-scale via ST. Thus, the background signal would be reduced in each subband. Secondly, the fourth partial differential equation method is adopted to modify the coefficient of each sub-band, which is for background suppression and anomaly signal enhancement. Thirdly, the kernel-based RX algorithm is adopted to detect the anomaly in each sub-band. Finally, the anomaly signal image is achieved by reconstructing the image with all modified sub-band. Several experiments with a HYDICE data are fulfilled to validate the performance of the proposed method. Compared with the original RX algorithm, experimental results show that the proposed algorithm has better detection performance and lower false alarm probability.

A new image registration method for infrared images and visible images

A new image registration method for infrared images and visible images is proposed in this paper. This method is a modified version of SIFT algorithm. Salient points of infrared images and visible images are extracted along edges of images. Each salient point is assigned a dominant direction which is the peak of orientation histogram of local image around it and a feature vector which is formed from the orientation histograms of sub-region around it. After salient points matching, RANSAC algorithm is used to eliminate wrong corresponding pairs. Experimental results show that this method has a good registration result for infrared images and visible images.

Multi-sensor image registration based on visual attention

Inspired by the process of manual registration, a method based on visual attention is proposed in this paper for multi-sensor image registration. In the first stage, the corner points are selected from both of the multi-sensor images using multi-scale Harris detector and then the outlines are extracted by Gabor filter. In the second stage, the selected points are described based on the contour images to find the matching pairs. Finally, the parameters of the affine transformation model between the images are obtained according to the matching pairs. Pairs of visible and infrared images are used to evaluate the performance of the proposed algorithm and SIFT algorithm. Experimental results show that the proposed method can achieve good performance for registering visible and infrared images.

A new non-uniformity correction algorithm for IRFPA based on statistical properties of scene

Influenced by detectors’ material, related manufacturing technology etc, every detection element’s responsivity in infrared focal plane arrays(IRFPA) is different, which results in non-uniformity of IRFPA. So non-uniformity correction(NUC) is an important technique for IRFPA. The classical two-point NUC algorithm based on reference sources is analyzed in this paper. And a new NUC algorithm based on statistical characteristics of image serial is presented. In this algorithm, the reference images are constructed from image serial, and correction parameters are computed by using the constructed reference images. Then two-point NUC is applied to output images of IRFPA. Experimental results show that the algorithm proposed in this paper is effective and implemented easily.

Infrared decoys recognition method based on geometrical features

After a decoy is released, it can fly around the target aircraft in a short period of time. And it can radiate infrared spectral radiation similarly to the target do. So it is difficult to recognize the target aircraft. But in infrared images, decoys and targets have different geometrical features. So an infrared decoys recognition method based on the geometrical features is proposed in this paper. The geometrical features of the candidates in each image are extracted, such as the major axis, the minor axis, the aspect ratio, area etc. Then the differences on these geometrical features can be used to recognize targets and decoys. A simulation was done on a set of images that contain decoys and targets by using this method. The results show that the algorithm proposed in this paper can better distinguish infrared decoys and targets.

Image Registration Algorithm Based on Modified GLOH Descriptor for Infrared Images and Electro-Optical Images

A new image registration algorithm for infrared images and electro-optical images is proposed in this paper. This algorithm is a combination of SIFT feature extraction algorithm and GLOH feature descriptor. Salient points of infrared images and electro-optical images are extracted along edges of images by using SIFT feature extraction algorithm. Then each salient point is described by using modified GLOH descriptor that formed a feature vector from the orientation histogram of sub-region around each salient point. After salient points matching by using Euclidean distance, RANSAC algorithm is used to eliminate wrong corresponding pairs. Then registration of infrared images and electro-optical images is achieved by affine transformation and bilinear interpolation. Experimental results for registration of infrared images and electro-optical images show that this algorithm has a good registration result.