Summary: | 碩士 === 國立高雄第一科技大學 === 電腦與通訊工程研究所 === 105 === In our daily life, digital images are often captured in the situation that light-source locates behind the object. In this case, the foreground objects are very dark and the image qualities are poor. Therefore, it is desirable to develop various backlighting image enhancement algorithms to increase the contrast of dark object and improve the visual quality. In this thesis, three backlighting image enhancement methods are presented. First, the enhancement method based on color saturation is developed. Four color saturations defined in color spaces IHS, HSI, HSV and HSL are investigated. The multiple-exposed images are generated and fused to obtain the final enhanced image. Second, the matrix homomorphic filtering method is studied. This method is based on the illumination-reflectance model and performs high-pass filtering in the logarithmic domain. Third, the dark channel prior and rank order filter are used to develop the enhancement method. The dark channel prior is used to generate multiple-exposed images, then they are fused to obtain the final enhanced image. Several backlighting images are used to evaluate the performance of the above three enhancement methods. Experimental results demonstrate that image contrast and quality can be improved by proposed methods.
|