Image Matching Process on Images Taken by a Cellular Phone
Yi Hu, Tomoharu Nagao, Masanori Okazaki, Taishi Chinone · The Journal of The Institute of Image Information and Television Engineers · 2003
With the popularity of cellular phones featuring built-in cameras, it is no surprise that content for information transfer using images taken from these phones is in great demand. Therefore, the need for correspondent image processing technology and a system to support content development is becoming much more urgent. A new method of image matching, which contains two approaches, is described. The first approach is to segment objects from images having a relatively simple background by applying a method for growing recursive regions and for the matching. The other is to do matching directly by using a presented GA algorithm in images with complex backgrounds to extract remarkable characters such as magazine titles and brand names on commodities, which are difficult to be segmented from their environment using existing methods. Experimental results show that the approaches work effectively in our system.