Infrared Image Enhancement based on Saliency Weight with Adaptive Threshold

Raw infrared images suffer the problem of low contrast. In this paper, a novel saliency weight based infrared image enhancement method is proposed. For the purpose of preserving the relative thermal distribution information, global-mapping enhancement is preferred in this work. To address the problem of under-enhancement on object areas in probability-histogram based methods, the spatial saliency information is utilized to develop a new weight of intensity levels for global-mapping to take the place of occurrence probability, named saliency weight. In the saliency weight, the weight values of objects-related intensity levels are larger than the values of background-related intensity levels, providing the basis to enhance objects. Based on the saliency weight, the concept of object-salient contrast is proposed to measure the contrast between objects and background. Enhancement is modeled as an optimization of maximizing the object-salient contrast. An adaptive threshold is established for the saliency weight to modify the solution of optimization, which prevents the background from being over-suppressed. Image intensities are re-distributed based on the solution through global-mapping. Experimental results demonstrate that the proposed method obtains larger contrast between objects and background than traditional probability histogram based methods after enhancement.