Analysis of methods of image feature extraction

Abstract

The article considers the methods of extraction and representation of image features, which play an important role in the processing of multimedia data. An overview of the latest developments in the field of image feature extraction is performed: extraction of color features from images where color has many functions, such as color moments(CM), color coherence vector (CCV), etc., extraction of texture features using the Gabor filter and extraction of shape features using methods based on contours and regions. An overview of the methods for representing image features is performed. The efficiency of merging global and local features in image processing was analyzed, as well as another way of presenting features like a visual word package. In the course of further work, the authors plan to study the relationship between the number of functions and the final performance. Intuitively, the larger the number of functions, the better the final performance. Secondly, the study of the relationship between the representation of functions and the final performance is also a very interesting and complex topic. It includes methods for representing objects (global, block, and regional). In the case of using block and regional functions, it is necessary to investigate the performance, which depends on the size of the partition or segmentation. Third, it is supposed to investigate the relationship between a suitable combination of image functions and the final performance in order to assess whether this combination can further improve performance. The results of the research are supposed to be used in the geolocation system of photos based on the automatic analysis of their images. Improving the accuracy and performance of image feature extraction will help improve the quality of the geopositioning system proposed in the article [28].