Image layer separation refers to one series of problems that try to decouple the input image into different component layers. While it has many practical applications, it is very challenging to solve due to its ill-posedness nature. Usually additional information and assumptions are required to make the problem tractable. In this thesis we focus on three types of specific layer separation problems – reflection separation, intrinsic images decomposition, and haze removal. We try to find our solutions to these problems by using different information or imposing new constraints. We first describe our approach for reflection removal, by using a small set of input image captured from different viewpoints. Next a layer separation method for single image input is presented with application to both reflection removal and intrinsic images estimation. The third work targets on the specific problem of recovering the visibility in nighttime haze scene. Besides these three layer separation works we also describe an application of layer separation that aims to remove compression artifacts raised in image contrast boosting. In all four works, our solutions show excellent performance in experimental results that are either on par with or superior to the current state-of-the-art. A summary chapter is included to summarize our contributions. In the end of the summary chapter, potential future directions are also presented.