基于颜色方差的图像相似检索算法研究

Research on Image Similarity Retrieval Algorithm based on Color Variance

针对均值哈希算法中存在的图像内容信息严重丢失、图像特征难以区分等造成的检索准确度不高的情况,提出了基于颜色方差的图像相似检索算法。算法分别对行元素、列元素进行灰度值方差求解,生成图像的横向方差向量和纵向方差向量,再通过余弦定理对其进行相似度计算。实验结果表明,算法在图像相似检索中执行效率较均值哈希算法提升了25.4%左右,且较均值哈希算法、感知哈希算法、差异值哈希算法能够更准确、完整地检索所有相似图像,具体检索完整性为"本算法>感知哈希算法>差异值哈希算法>均值哈希算法"。

Aiming at the fact that the image content information in the mean hash algorithm is seriously lost and the image features are difficult to distinguish, the image similarity retrieval algorithm based on color variance is proposed. First, the algorithm separately solves the variance of the gray value of the row elements and column elements, and generates the horizontal variance vector and the longitudinal variance vector of the image. Then, the similarity calculation is performed by the cosine theorem. The experimental results show that the algorithm performs better than the mean hash algorithm in image similarity retrieval by about 25.4%, and the average hash algorithm, perceptual hash algorithm and difference value hash algorithm can retrieve all similar images more accurately and completely. The experimental results indicate that the efficiency of the algorithm in image similarity retrieval is about 25.4% higher than that of the average hash algorithm. Compared with the mean hash algorithm, the perceptual hash algorithm and the difference value hash algorithm, the proposed algorithm can retrieve all similar images more accurately and completely. The specific retrieval integrity is "proposed algorithm> perceptual hash algorithm> difference value hash algorithm>mean hash algorithm".