Self-adaptive Luminance- And Chrominance- Balance Image Fusion Algorithm Based On Local Features

With the fast development of sensor and image processing techniques, the image fusion has become one of the hot topics in information fields, and its applied potential has got sufficient attention and recognition. On the one hand, image fusion has been widely applied in military field including targets detection, identification and tracking. On the other hand, in the field of the remote sensing, medical diagnosis, industrial testing, intelligent robot, machine vision and the application of digital camera, image fusion has got extremely widespread application prospects.In nearly 30 years, people have done a lot of theoretical research and practical application on the different levels of image fusion. A number of traditional image fusion methods have been proposed in the early stage, which mainly include averaging algorithm of the pixel values, image fusion method based on the pyramid and wavelet transform, the method based on strategies of region segmentation and so on. In these methods, algorithms based on the pyramid and wavelet transform belong to the image fusion method which is based on multi-scale decomposition, the main idea of which is that transforming the registered source images from spatial domain to frequency domain by wavelet or pyramid transform. Then conduct image fusion using some fusion strategies on the basis of the transformed images. At last, desired image is constructed by performing an inverse transform. Image fusion strategies based on wavelet transform is superior to the standard methods in terms of minimizing color distortion and reducing the loss of contrast information. Along with the further research in recent years, many algorithms combining several fusion strategies are proposed, such as, Multi-focus Image Fusion Based on Image Block Segment and Wavelet Space Frequency, which can obtain better results than either wavelet transform or image block method alone. For the two neural image fusion algorithms for color and gray level images, it's implementation for both color and gray level images shows that the quality of noisy images can be enhanced effectively. As one of the importance branch in image fusion, the multi-focus image fusion is widely applied in the field of the computer vision, digital camera clear imaging and target detection. For multi-focus images, this paper presents a self-adaptive luminance- and chrominance-balance image fusion algorithm based on local features. Firstly, in order to extract local features of images, it used curvature estimation approach based on parabola fitting. Thus, weight calculation models are established by analyzing the local features, which can automatically get clear image pixels. Finally, this paper adjusted brightness and chrominance of images by adopting the idea of multi-band fusion, and thus, the smooth fused image is obtained. The experimental results show that this algorithm can fuse several images taken under different focus levels, eliminate the difference of lighteness and chrominance in the images, and achieve smooth transition between images.