Abstract: An observed image which is characterized not only contains the geometric structure, but also may contain noise or texture. Image feature classification and segmentation play an important role in the fields of image processing and pattern recognition (such as de-noising, edge detection and texture segmentation). From the function space and frequency-domain transformation, this paper introduces the represent methods of image feature and compares it, summarizes the advantages and disadvantages of each. At the end, the application area of image feature study is given in personal opinion.

Key words: image feature, function space, bounded variation, wavelet transform, contourlet transform