E.g. similar images that have identical values for a small number of relevant features may nevertheless be distant from one another in the feature space (due to other irrelevant features) and may therefore be judged by machines as dissimilar using the simultaneous method. Hence, the importance of different image features in image characterization or similarity assessment should not be equal. If we select the relevant or salient features at a proper hierarchical level and those irrelevant features to a very low level, and only check these low level features when necessary, we may avoid the curse of dimensionality. In other words, if we select the features correctly in building up an image characterization hierarchy, we may improve the performance of similarity assessment.帮我翻译一段文字,感激不尽
解决方案 »
免费领取超大流量手机卡,每月29元包185G流量+100分钟通话, 中国电信官方发货