A Study on Image Enchantment Algorithm for Exposure Problems

• Main Authors: Min-Chueh Kan, 甘旻爵
• Other Authors: Yu-Xiang Zhao
• Format: Others
• Language: zh-TW
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/48744272431137228041

碩士 === 國立金門大學 === 理工學院工程科技碩士在職專班資訊系統組 === 104 === With the advance of technology and science, more and more people use smartphones and digital cameras to shoot their outdoor life photos. However, the outdoor life photos usually have the exposure problems due to the influence of light. These problems can be summarized as photo underexposed and photo overexposed. Histogram equalization (HE) is the most common algorithm used to enhance the photo exposure problems. But this algorithm is based the global brightness enhancement which has the photo over-enhance problem. For this reason, we propose two algorithms, picture exposure enhancement algorithm (PEEA) and genetic algorithm based picture exposure enhancement algorithm (GA-PEEA), to enhance the photo exposure problems. In the PEEA, the k-means algorithm is used to divide the histogram into two parts in the proposed PEEA: dark part and bright part. The clustering parameters are used to design the transfer function of the exposure enhancement, and the transfer function is used to enhance the original picture. The proposed PEEA will tune the parameters of the transfer function and evaluate the effectiveness of our proposed evaluation of picture exposure level (EPEL). The transfer function of the best EPEL is tuned by PEEA and the transfer function is used to enhance the exposure picture. In the GA-PEEA, the genetic algorithm is used to find the optimal EPEL of the transfer function and the transfer function is used to enhance the exposure picture. In the experiment, the proposed algorithms are evaluated by mean square error (MSE), signal to noise ratio (SNR) and peak signal to noise ratio (PSNR), and compared with HE, brightness preserving bi-histogram equalization (BBHE) and dualistic sub-image histogram equalization (DSIHE). Finally, the experimental results show that the proposed algorithms are better than HE, and achieve satisfied performance than BBHE and DSIHE.