Font Size: a A A

Research On Cast Correction And Shadow Removal In Color Images

Posted on:2011-10-09Degree:MasterType:Thesis
Country:ChinaCandidate:H YuFull Text:PDF
GTID:2178330338981669Subject:Probability theory and mathematical statistics
Abstract/Summary:PDF Full Text Request
Illumination correction is one hot topic in low-level computer vision fields. Color cast correction and shadow removal, two challenge problems of illumination correction, are discussed in this thesis.In color cast correction, several representative color spaces are introduced firstly, then color constancy theory and classical color cast correction techniques are discussed. After that, a novel gray-axis adjustment based color cast correction algorithm is pre-sented. The key of color cast correction is the procedure of light conditions estimation. In the proposed algorithm, the statistical characteristics of the original color image are first computed to estimate the light conditions. Then, the actual gray-axis is fitted using PCA. Finally, color correction is achieved by a specific orthogonal transformation in CIELab color spaces. Experimental results demonstrate that the proposed algorithm can effectively correct color cast under the uniform illumination conditions.In shadow removal, firstly the shadow models and related background knowledge are introduced, including the concepts of umbra and penumbra, as well as traditional shadow detection and removal methods. Then, the illuminant invariant theory and related methods proposed by Finlayson are discussed. By analyzing the deficiency of Finlayson's method, an improved scheme is presented, which can maintains the texture information in the process of shadow removal. Two important steps of the proposed algorithm are as follows:(1) S-shaped curve are employed to estimate the illumination variance of penumbral region, and the corresponding inverse transform is performed to eliminate unexpected impacts caused by penumbral region; (2) as for umbra region, a specific color correction method is adopted, which reconcile the umbra region with the lit region in statistics characteristics (such as mean, variance, etc). Experimental results show that the processed images not only maintain the texture information of the penumbra region, but also have consistent texture of the umbra region and the lit region.
Keywords/Search Tags:color image pre-processing, color constancy, color cast correction, shadow removal
PDF Full Text Request
Related items