The quadratic-chi histogram distance family

摘要: We present a new histogram distance family, the Quadratic-Chi (QC). QC members are Quadratic-Form distances with cross-bin χ2-like normalization. The normalization reduces effect of large bins having undo influence. Normalization was shown to be helpful in many cases, where χ2 outperformed L2 norm. However, is sensitive quantization effects, such as caused by light changes, shape deformations etc. part takes care relationships (e.g. red and orange), alleviating problem. two crossbin properties: Similarity-Matrix-Quantization-Invariance Sparseness-Invariance show that have these properties.We also experimentally they boost performance. computation time complexity linear number non-zero entries bin-similarity matrix histograms it can easily parallelized.We results for image retrieval using Scale Invariant Feature Transform (SIFT) color descriptors. In addition, we classification Shape Context (SC) Inner Distance (IDSC). outperform state art tasks, while short running time. experimental both property important.