Research On The Fusion Algorithm Of Infrared Image And Visible Image

Image fusion is a popular research topic to generate a composite image by integrating the complementary information from multiple source images, which are obtained by multiple sensors of the same scene. The information of infrared image is decided by object and scenes radiance rate and temperature difference,which the targets could be easily find,However, the infrared image has lower contrast and weak details on the other hand,the visible sensor can obtained more detail information of the scene.So the fusion method of infrared and visible images has widely employed in many applications such as military detection/remote sensing.Multi-scale transform(MST) theories are one of most popular tools among pixel level fusion methods. For some specific MST, we compared the effectiveness of some specific MST using the objective fusion measure, the impact of the decomposition level and different fusion rules are studied experimentally.Currently sparse representation(SR) gains considerable interest and is applied in many fields. Sparse representation addresses the signals’ natural sparsity,which is in accord with the physiological characteristics of human visual system, As the dictionary is over-complete, the SR methods can achieve more meaningful representations of the source images. The detailed fusion scheme based on SR are proposed. Our experimental results have indicated that SR-based methods has better fusion performance in some noiseless and noise situation.In the end,the advantages and disadvantages of MST-methods and SR-methods are fully analyzed theoretically.we proposed an image fusion framework combining the MST and SR methods. The registered original images were decomposed into the low-frequency and high-frequency coefficients by MST transform. The sparse coding technique is employed for the fusion of MST low-frequency. High-frequency coefficients are fused using the popular “absolute-values-max” rules. Then we can obtained the fused image through the inverse MST. The experiment can prove that this fusion framework have a better performance over conventional MST-and SR-based methods.