Multi-characteristics Image Retrieval
Yinan Zhang, Wenming Cao · Advances in Social Science, Education and Humanities Research/Advances in social science, education and humanities research · 2015
In this paper, according to different images of shifting, rotation and scale invariance properties, we propose a multiple features method for image retrieval.Firstly, we get the main area of the image to avoid the affections by some nonsignificant scenes.Secondly, we split the image into five parts, and utilize the pixels distribute information of every sub-image to get the representative points.Positional relationship among these representative points of the sub-blocks remain the same when the image suffers shifting or scaling transformation.By this means, the image can be expressed in a sample way and avoid complex shape detection.Then, in order to get the more detail information, the image is divided into several concentric rings, and we statistic the number of pixels in each annulus area to construct the third feature.Finally, combining these features with a certain weight coefficients and the hybrid characteristic is put forward.Experiments show that this method can achieve good effect of image retrieval and flexible to compute.