Research And Application Of Texture Feature Extraction Algorithm

Texture is the most important way of visual perception,and it is also a major way for human beings to distinguish different objects.In recent years,texture has become a hotspot in image processing.Texture-based image retrieval is an important direction in the field of image processing.In the retrieval process,an efficient texture feature extraction algorithm is significant for the improvement of user experience and retrieval effect.In order to verify the high efficiency of the texture feature extraction algorithm,a retrieval system based on texture is designed and implemented,and the texture feature extraction algorithm is improved.First,the traditional Edge Histogram Descriptor(EHD)operator will loss part of information of the image in the division.For the shortcomings,a new idea is proposed,choosing to cut the original image before dividing it,which usually lost the outside useless information.Secondly,the Local Binary Pattern(LBP)algorithm can extract the details of the image,while Gabor wavelet can extract the texture in multi-direction and multi-scale.A new mind named Gabor-LBP is proposed through combining the two algorithm.In the method,on the basis of Gabor wavelet,the image is rotated invariant LBP processing,and the texture details are extracted.Finally,the improved EHD algorithm is combined with the Gabor-LBP algorithm in a linear way,and the combined algorithm is applied to image retrieval to extract both the image edge texture and the image texture detail.The response time,precision and recall of the image retrieval system are used to evaluate the texture feature extraction algorithm.The experimental results show that the improved EHD is better than the original EHD,and the Gabor-LBP algorithm is better than the single Gabor filter.In the case of improving EHD and Gabor-LBP combined in a linear way,when the weight of the improved EHD is 0.6,the retrieval effect is the best.And the accuracy of combined algorithm is obviously higher than that of EHD and Gabor-LBP itself.