There are many kinds of corporeal ingredients in urine sediment which must be identified to confirm the diagnosis of an abnormality. With the development of image processing, pattern recognition and machine learning, automatic identification is a new trend. In this thesis, we will give a simply introduction about auto- indentifi- cation in urine sediment and some basic concepts of image processing and pattern recognition. And then we refine a method which integrates both Gabor filter and scattering transform for texture analysis in urine sediment images. The proposed scheme is based on the techologies, conventional Gabor filter and the recently developed scattering transform used in texture extraction. The Gabor filter bank has the ability to capture the filtering responses according to the scale and orientation of texture. Besides, the scattering transformation provides a distinctive property of robust description, which is invariant to rotations and stable to spatial deformation. The excellent representation of Gabor filter and scattering transform has been severally studied in recent work, yet they have not been used in urine sediment images. In this work, we propose to use both Gabor filter and scattering transformation to extract the texture feature of urine sediment images. Coupling with an efficient support vector machine (SVM) classifier, the proposed scheme tends to shown superiority as compared to other single descriptive alternatives in real urine sediment experiments.