An improved detail enhancement method for colorful image via guided image
Yunlan Tan, Taozhi Si, Guangyao Li, Mang Xiao · 2014
In this paper, we propose an improved detail enhancement method via guided image. To achieve the best overall balance for the contradictory goals of edge-preserving smoothing and details capturing, we propose a method combining the guided image filter (GIF) with local detail enhancement and the weighted least squares (WLS) filter with global intensity shift for input image. To do so, we first make edge-preserving smoothing operators in the R, G, B channels respectively by the guided filter. We then utilize the former enhanced image as a model and execute global enhancement in the luminance channel. The important contents are well preserved without distorting the overall image structure which does not suffer from the gradient reversal artifacts in detail enhancement due to the abrupt change of the edge. This method can produce high-quality detail enhancement and edge-preserving smoothing. In addition, it produces halo-free edge-preserving smoothing because it distributes blurred edges globally. Experiments show that the improved guided filter is both effective and efficient in a great variety of computer vision and computer graphics applications, including edge-aware smoothing, detail enhancement, etc.