چكيده فارسي :
ﭼﮑﯿﺪه: در ﺗﺼﻮﯾﺮ ﺑﺮداري از ﯾﮏ ﺻﺤﻨﻪ در ﺻﻮرﺗﯽ ﮐﻪ ﻋﻤﻖ اﺷﯿﺎء ﻣﻮﺟﻮد در ﺻﺤﻨﻪ از ﻋﻤﻖ ﻣﯿﺪان ﻟﻨﺰ ﺑﯿﺸﺘﺮ ﺑﺎﺷﺪ، ﻓﻘﻂ اﺷﯿﺎء و ﯾﺎ ﻗﺴﻤﺘﻬﺎﯾﯽ از اﺷﯿﺎء ﮐﻪ در ﻣﺤﺪوده ﻋﻤﻖ ﻣﯿﺪان ﻫﺴﺘﻨﺪ واﺿﺢ ﻣﯽﺷﻮﻧﺪ و ﺑﻘﯿﻪ ﻗﺴﻤﺘﻬﺎ ﻣﺎت ﺧﻮاﻫﻨﺪ ﺷﺪ. ﺟﻬﺖ ﻏﻠﺒﻪ ﺑﺮ اﯾﻦ ﻣﺸﮑﻞ از ﺗﮑﻨﯿﮑﯽ ﺑﻪ ﻧﺎم ادﻏﺎم ﺗﺼﺎوﯾﺮ اﺳﺘﻔﺎده ﻣﯽﺷﻮد. در اﯾﻦ ﺗﮑﻨﯿﮏ ﺑﺎ ﺗﻐﯿﯿﺮ ﺗﻤﺮﮐﺰ دورﺑﯿﻦ ﭼﻨﺪ ﺗﺼﻮﯾﺮ اﺧﺬ ﻣﯽﺷﻮد ﮐﻪ در ﻫﺮﯾﮏ، ﺑﺨﺸﯽ از ﺻﺤﻨﻪ واﺿﺢ ﺧﻮاﻫﺪ ﺑﻮد. ﺳﭙﺲ اﯾﻦ ﺗﺼﺎوﯾﺮ )ﮐﻪ ﺑﻪ آﻧﻬﺎ ﺗﺼﺎوﯾﺮ ﭼﻨﺪﺗﻤﺮﮐﺰي ﮔﻔﺘﻪ ﻣﯽﺷﻮد( ﺑﺎ ﯾﮑﺪﯾﮕﺮ ادﻏﺎم ﻣﯽﺷﻮﻧﺪ ﺗﺎ ﯾﮏ ﺗﺼﻮﯾﺮ ﮐﻪ ﻫﻤﻪ ﺑﺨﺸﻬﺎ واﺿﺢ ﺑﺎﺷﻨﺪ ﺑﻪ دﺳﺖ آﯾﺪ. در اﯾﻦ ﻣﻘﺎﻟﻪ، ﯾﮏ روش ادﻏﺎم ﺗﺼﺎوﯾﺮ ﭼﻨﺪﺗﻤﺮﮐﺰي ﮐﻪ ﺗﻐﯿﯿﺮ ﯾﺎﻓﺘﮥ ﯾﮏ اﻟﮕﻮرﯾﺘﻢ ﻣﻮﺟﻮد اﺳﺖ ﭘﯿﺸﻨﻬﺎد ﺷﺪه اﺳﺖ. در روش ﻣﻮﺟﻮد، اﺑﺘﺪا ﻧﻮاﺣﯽ ﺑﺮﺟﺴﺘﻪ در ﻫﺮ ﺗﺼﻮﯾﺮ ﻣﻨﺒﻊ، ﺗﺸﺨﯿﺺ داده ﺷﺪه و ﺳﭙﺲ ﺑﺎ ﻣﻘﺎﯾﺴﮥ ﺑﺮﺟﺴﺘﮕﯽﻫﺎي ﻣﻮﺟﻮد ﺑﯿﻦ ﺗﺼﺎوﯾﺮ ﻣﻨﺒﻊ، ﻧﻮاﺣﯽ ﻣﻮرد ﺗﻤﺮﮐﺰ ﻣﺸﺨﺺ ﻣﯽﺷﻮﻧﺪ. در روش ﻣﺬﮐﻮر، از داﻣﻨﮥ ﺗﺒﺪﯾﻞ ﻓﻮرﯾﻪ اﺳﺘﻔﺎده ﺷﺪه و ﭘﺮدازﺷﯽ روي اﻃﻼﻋﺎت ﻓﺎز اﻧﺠﺎم ﻧﺸﺪه اﺳﺖ. در اﯾﻦ ﻣﻘﺎﻟﻪ، روﺷﯽ ﺑﺮﮔﺮﻓﺘﻪ از ﻣﻘﺎﻟﻪاي دﯾﮕﺮ ﺟﻬﺖ اﺳﺘﻔﺎده از اﻃﻼﻋﺎت ﻓﺎز ﻣﻮرد ﺑﺮرﺳﯽ ﻗﺮار ﮔﺮﻓﺘﻪ اﺳﺖ. در روش اﺳﺘﻔﺎده ﺷﺪه، اﺑﺘﺪا ﭼﻨﺪ ﺗﺼﻮﯾﺮ ﺟﺪﯾﺪ )ﮐﻪ ﺑﻪ آﻧﻬﺎ ﻧﻘﺸﻪﻫﺎي وﯾﮋﮔﯽ ﮔﻔﺘﻪ ﻣﯽﺷﻮد( ﺑﺎ ﺗﺮﮐﯿﺐ ﮐﺎﻧﺎﻟﻬﺎي رﻧﮓ ﺗﺼﺎوﯾﺮ ورودي اﯾﺠﺎد ﻣﯽﺷﻮد. ﺳﭙﺲ اﻃﻼﻋﺎت ﻓﺎز از ﻧﻘﺸﻪﻫﺎي وﯾﮋﮔﯽ در داﻣﻨﻪ ﻓﺮﮐﺎﻧﺲ اﺳﺘﺨﺮاج ﺷﺪه و از آﻧﻬﺎ ﺟﻬﺖ ﺗﻌﯿﯿﻦ ﻧﻮاﺣﯽ ﺑﺮﺟﺴﺘﮥ ﻫﺮ ﺗﺼﻮﯾﺮ و ﺳﭙﺲ ادﻏﺎم اﺳﺘﻔﺎده ﻣﯽﺷﻮد. ﺑﺎ ﺗﺴﺖ روش ﭘﯿﺸﻨﻬﺎدي روي ﯾﮏ ﻣﺠﻤﻮﻋﻪ از ﺗﺼﺎوﯾﺮ ﭼﻨﺪ ﺗﻤﺮﮐﺰي و ﻣﻘﺎﯾﺴﻪ ﺑﺎ ﭼﻨﺪ روش ﻣﻮﺟﻮد و ﻫﻤﭽﻨﯿﻦ روش ﭘﺎﯾﻪ، ﻧﺘﺎﯾﺞ ﺑﻬﺘﺮي ﺑﻪدﺳﺖ آﻣﺪ.
چكيده لاتين :
In imaging a scene, it is possible that the depth of the objects in the scene is larger than the depth of the field of the lens. In such cases, just objects or parts of objects that are in depth of field of the lens, will be sharp in the image while others become blurred. To overcome this problem, a technique called image fusion is used. In this technique, by changing the focus of the camera, some images are acquired such a way that in each image, part of the scene is sharp. Afterwards, these images (called multi focus images) are fused in a single image such that all parts be sharp. In this paper, a multi-focus image fusion method is proposed. The proposed method is based on an existing method which uses comparison of spectrum of the Fourier transform. In the existing method, first, salient regions are detected in each input image. Then, by comparing salient regions of input images, a focus map is extracted. In mentioned method, the amplitude of Fourier transform is processed while no process is performed on the phase angle. In this paper, a method taken from another paper is employed to benefit the phase information of the Fourier transform. In the proposed method, firstly, three images (called feature maps) are created by combining color channels of each input image. Afterwards, phase information is extracted from feature maps in frequency domain. Using them, salient parts of the images are determined and finally the fused image is produced. By testing the proposed method on some images, the results improved in comparison with the base method and some other methods.
