Pixel-based Multi-focused Image Fusion using Color Appearance Model

• Main Authors: Wen-hao Wu, 吳文豪
• Other Authors: Chen-Chiung Hsieh
• Format: Others
• Language: zh-TW
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/54411509888283802338

碩士 === 大同大學 === 資訊工程學系(所) === 103 === The technology of multi-focus imaging is to fuse multiple images, which have the same scene but with difference focal distance, into one picture with several focusing simultaneously. General multi-focus image fusion technology can only process gray image mostly. In this paper, it analyzes and processes color image and uses color saturation value that is based on Chrominance value divide to Luminance value directly, and it does not measure the focusing with traditional rectangle window mask. And when we analyze most of the focusing pixels, their focusing shapes are similar to star light shapes. Therefore, we name the focusing share Star-light focusing detection, which is used to measure the focusing status of the pixels. In addition, the minimum measurement unit is based on pixels in this discussion; pixel-based fusion can take more details than block-based or region-based fusion. And each pixel value on the fused image shall be the same as the pixel value on original image possibly. Therefore, it will be presented faithfully to the original image. Finally, considering actual application factors in photographing process, such as hand vibration, image distortion caused by optical assembly, wind blowing of the objects in the scene and so on, which make the same object on two original images have displacement or deformation phenomena, when processing fusion, it must consider registration. To verify resultant fused image, we shall use manual method to fuse a mask image and take it as reference to produce the best resultant fused image and calculate PSNR. Furthermore, this paper makes comparison between the fused finial mask image and synthetic fused mask image to calculate the correction rate and error rate. The paper also makes comparison between the two groups of color images produced by current method (state of the art). The obtained mean correction rate and PSNR are 80% and 42dB respectively. Additionally, two other groups of gray images processed by the obtained mean correction rate and PSNR are 83% and 40dB respectively. Method in this paper is superior to most referenced methods. It is proved that Color Saturation and Star-light focusing detection can be applied to image fusion reliably.